ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2017

Digital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Medicine 1309

Early Mortality After Total HipArthroplasty In Sweden

ANNE GARLAND

ISSN 1651-6206ISBN 978-91-554-9842-9urn:nbn:se:uu:diva-316989

Dissertation presented at Uppsala University to be publicly examined in Linnésalen, MuseumGustavianum, Akademigatan 3, Uppsala, Saturday, 29 April 2017 at 13:00 for the degree ofDoctor of Philosophy (Faculty of Medicine). The examination will be conducted in English.Faculty examiner: Professor Ashley Blom (University of Bristol, UK).

AbstractGarland, A. 2017. Early Mortality After Total Hip Arthroplasty In Sweden. DigitalComprehensive Summaries of Uppsala Dissertations from the Faculty of Medicine 1309.63 pp. : Uppsala University. ISBN 978-91-554-9842-9.

Every year 16 000 individuals receive a total hip arthroplasty (THA) in Sweden. Even thoughTHA is a common procedure, adverse events do occur. The most dramatic complication is deathin the postoperative phase. The overall aim of this thesis was to describe and investigate earlymortality after THA in Sweden.

Sweden has an ideal platform for national observational registry studies, thanks to the useof personal identity numbers. Operation-specific information was collected from the SwedishHip Arthroplasty Register, medical information from the National Board of Health and Welfare,and socioeconomic information was collected from Statistics Sweden. Main outcome was 90-day mortality.

Study I was a prospective observational register study investigating the risk of mortality aftera simultaneous bilateral THA compared with staged bilateral THA. There was no clinicallyrelevant difference in early postoperative mortality between the two groups.

Studies II and III were nation-wide matched cohort studies, with adjustment for comorbidityand socioeconomic background. Adjusted early mortality in femoral neck fracture patientsreceiving a THA is about double compared with a matched control population. Young (60-69years) femoral neck fracture patients receiving a THA have a low absolute mortality risk, whilethose who are older than 80 years with a higher degree of medical comorbidity run a high risk ofearly death (II). In study III healthier, younger patients with higher socioeconomic status tendedto be selected for cementless THA, resulting in selection bias. Even after accounting for thisbias, however, there remains a small absolute and adjusted increase in the risk of death within14 days after elective THA surgery using fully cemented implants.

Study IV was a nationwide prospective cohort study comparing different comorbiditymeasures in terms of predicting early postoperative mortality after THA. A less data-demandingcomorbidity measure is better at predicting 90-day mortality than more commonly used codingalgorithms.

In conclusion, socioeconomic background and the presence of comorbidities have animportant influence on early mortality after THA, while the type of fixation is of less importance.Future mortality studies could benefit from the use of data that are routinely collected, and thusavoid the logistically complicated procedure now necessary to merge national databases.

Keywords: early mortality, total hip replacement, total hip arthroplasty, comorbidity,socioeconomic factor, cemented

Anne Garland, Department of Surgical Sciences, Orthopaedics, Akademiska sjukhuset,Uppsala University, SE-75185 Uppsala, Sweden.

© Anne Garland 2017

ISSN 1651-6206ISBN 978-91-554-9842-9urn:nbn:se:uu:diva-316989 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-316989)

Populärvetenskaplig sammanfattning

Varje år får ca 16 000 individer “en ny höft” i Sverige. Det innebär att led-skålen i bäckenet ersätts med en konstgjord skål, oftast tillverkad av plast, och ledhuvudet ersätts med ett metall- eller keramiskt huvud förankrat på en stam av metall. Höftproteskirurgi är att betrakta som rutinkirurgi idag men det är ett stort kirurgiskt ingrepp som inte är riskfritt för patienten. De van-ligaste orsakerna till att man får en höftprotes är artros eller brott på lårbens-halsen. Indikationerna för proteskirurgi har under åren vidgats vilket gör att fler både yngre och äldre patienter opereras än tidigare. Äldre patienter har av naturliga skäl en högre risk för allvarliga komplikationer medan de yngre som opereras förefaller oftare ha andra sjukdomar utöver höftsjukdomen. Idag opereras fler riskpatienter än tidigare.

I avhandlingen ingår fyra registerbaserade delarbeten där dödligheten ef-ter höftproteskirurgi har studerats. Information om protessort och operat-ionsdetaljer har hämtats från Svenska Höftprotesregistret, information om diagnoser, läkemedel och dödsorsaker från Socialstyrelsens patientregister, läkemedelsregister och dödsorsaksregister, och information om socioekono-misk bakgrund från Statistiska Centralbyrån. Datan har analyserats med avancerade statistiska metoder medan frågeställning (val av studiedesign) och tolkning av resultaten har en klinisk utgångspunkt. Död inom 90-dagar är ett utfallsmått som ofta används för att värdera risker med olika medi-cinska behandlingar. Orsakerna till att en patient avlider i samband med eller inom 90 dagar efter en höftledsoperation kan vara många men de domine-rande orsakerna borde vara hjärt-och kärlsjukdom, sjukdomar i hjärnans blodkärl eller blodproppssjukdom. Det huvudsakliga utfallsmåttet i ana-lyserna har genomgående varit död inom 90 dagar.

Samsjuklighet mäts ofta med komplexa index som beräknas utifrån vilka diagnoser som har registrerats. Dessa index är inte tillämpbara i den kliniska vardagen och ett väletablerat kliniskt riskvärderingsinstrument saknas. I studie fyra utvärderas olika samsjuklighetsmått och deras förmåga att förut-säga död inom 90 dagar efter höftprotesoperation. Med statistiska värde-ringsmetoder tar jag fram en kombination av klinisk lättillgänglig informat-ion som i framtiden skulle kunna utvecklas till ett riskvärderingsinstrument.

Sammanfattningsvis ger resultaten av delarbete I-IV följande: • 90-dagarsdödligheten efter höftprotes har ett medelvärde i riket på 7,1

‰ (det vill säga ca 7 dödsfall per 1000 opererade patienter). • Höftproteskirurgi på båda sidorna vid samma operationstillfälle ger inte

någon relevant skillnad i risken att dö inom 90 dagar jämfört med att operera en höft i taget.

• För patienter som opererats med höftprotes efter höftfraktur är risken att dö inom 90 dagar ungefär dubbelt så hög som för en matchad kontroll-grupp. Höftfrakturpatienter under 70 har en låg risk medan de över 80 år med multipla sjukdomar har en hög risk att dö inom 90 dagar efter höft-protesoperation.

• Såväl samsjuklighet i andra sjukdomar som socioekonomisk bakgrund har betydelse för risken att dö inom 90 dagar efter operation medan val av protesens fixationssätt har ett mer marginellt inflytande. Då cemente-rad protes används vid höftprotesoperation på grund av artros ses en li-ten höjning av dödligheten de första två veckorna jämfört med en icke opererad kontrollpopulation. Denna höjning motsvarar 5 dödsfall per 10 000 opererade patienter. Motsvarande höjning ses inte då ocementerad protes används.

• Studie IV visar att en att ett litet antal utvalda riskfaktorer tillsammans är bättre på att förutsäga död inom 90 dagar än mer komplexa index.

Mortality : the way that people do not live for ever

: the number of deaths within a particular society and within a

particular period of time

Definition from www.dictionary.cambridge.org

Prologue

Welcome to my book. I know that some of you who hold the book in your hands won’t actually read it. But since it might be the only book I will ever write I would like to give you a warm welcome. Then we will see how it goes. No pressure.

I often find myself reflecting about all those small things that lead to other small things that in turn, however insignificant, in the end might lead to something overwhelmingly important like “Life or Death” or even “a book or no book”. In early spring 2009, my last day as a visiting registrar doctor, I sat in a small, overcrowded office at Södersjukhuset in Stockholm with a breathtaking view over the city. Thanks to that view I didn’t mind too much that my designated mentor was more into talking about “pros and cons” of life in general than about my own “pros” and “still to develops”. We round-ed up the performance appraisal with a little small talk about our plans for the future. He was planning to retire. I was planning to return to Gotland. With our goodbyes said he stopped me when I already was more than half-way on my way out, looked at me sternly above his glasses and said:

- Dr Garland, you’re not planning to die “dissertation-less”, are you?*

That’s been bugging me ever since. It was a small thing, but it led to other small things that in the end led to this book. Welcome to my book “Early Mortality After Total Hip Arthroplasty In Sweden”.

* “Dr Garland, du tänker väl aldrig dö odisputerad?”

This thesis is the result of collaboration between the Institute of Surgical Sciences Uppsala University, Region Gotland

and the Swedish Hip Arthroplasty Register.

List of Papers

This thesis will be based on the following papers, which are referred to in the text by their Roman numerals.

I Garland, A., Rolfson, O., Garellick G., Kärrholm J., Hailer N.P. (2015) Early postoperative mortality after simultaneous or staged bilat-eral total hip arthroplasty: an observational register study from the Swe-dish Hip Arthroplasty Register. BMC Musculoskelet Disord., Apr 8; 16:77

II Hailer N.P., Garland A., Rogmark C., Garellick G., Kärrholm J.

Early mortality and morbidity after total hip arthroplasty in patients with femoral neck fracture. Acta Orthop. 2016 Sep 20:1-7.

III Garland A., Gordon M., Garellick G., Kärrholm J., Sköldenberg O.,

Hailer N.P. Risk of early mortality after cemented compared with ce-mentless total hip arthroplasty: A nationwide matched cohort study. Bone Joint J. 2017 Jan;99-B(1):37-43.

IV Garland A., Nemes S., Bulow E., Garellick G., Rolfson O., Hailer

N.P. Prediction of 90-day mortality after major surgery made simpler: An analysis of different comorbidity measures based on 38,735 patients from the Swedish Hip Arthroplasty Register. In manuscript.

Contents

Abbreviations ................................................................................................ 15

Introduction ................................................................................................... 17Total Hip Arthroplasty - “The operation of the century” ......................... 17THA- Different Fixation Methods ............................................................ 18Evaluating Mortality After Total Hip Arthroplasty .................................. 19Mortality After Bilateral Total Hip Arthroplasty ..................................... 19Mortality After Total Hip Arthroplasty Due To Femoral Neck Fracture . 20Comorbidity .............................................................................................. 20Socioeconomic Background ..................................................................... 23Sources of Epidemiological Data ............................................................. 23

The Swedish Hip Arthroplasty Register .............................................. 23Statistics Sweden ................................................................................. 24The National Patient Register .............................................................. 24The Drug Register ................................................................................ 24

Ethics and Legislation .............................................................................. 25

Aims .............................................................................................................. 27

Methods and Study Populations .................................................................... 28Ethics ........................................................................................................ 28Methods .................................................................................................... 28

Study I .................................................................................................. 28Study II ................................................................................................. 29Study III ............................................................................................... 29Study IV ............................................................................................... 30

Statistical Methods ................................................................................... 31

Results ........................................................................................................... 34Study I ....................................................................................................... 34Study II ..................................................................................................... 36Study III .................................................................................................... 38Study IV .................................................................................................... 40Further Results .......................................................................................... 42

Comorbidity- Trend over time ............................................................. 42Mortality- Trend over time .................................................................. 43

Discussion ..................................................................................................... 44Methodological Considerations ................................................................ 47

Conclusions ................................................................................................... 50

Future Research ............................................................................................. 51

Project Collaborators ..................................................................................... 52

Acknowledgments ......................................................................................... 54

References ..................................................................................................... 57

Epilogue ........................................................................................................ 63

15

Abbreviations

ASA American Society of Anesthesiologists CCI Charlson Comorbidity Index CI Confidence Interval DAG Directed Acyclic Graph HR ICD

Hazard Ratio The International Classification of Dis-eases

NARA Nordic Arthroplasty Register Association NPR National Patient Register/ Patientregistret RxRiskV Prescription-based comorbidity measure SMR Standardized Mortality Rate SHAR Swedish Hip Arthroplasty Register/ Sven-

ska Höftprotesregistret THA Total Hip Arthroplasty

16

17

Introduction

Total Hip Arthroplasty - “The operation of the century” Osteoarthritis (OA) is a painful, often disabling condition but not a threat to life itself. Non-invasive treatment alternatives include physiotherapy, learn-ing coping strategies including the use of walking aids, non-steroidal anti-inflammatory drugs and sometimes morphine. When non-invasive treatment no longer gives satisfactory pain relief a total hip arthroplasty (THA) be-comes an option. A successful THA can give the patient with OA in the hip a pain-free joint, the possibility to keep an active lifestyle without impair-ment, and an improved quality of life.67

Sir John Charnley made an invaluable contribution to the treatment of hip OA by introducing the THA.13 The THA is a surgical procedure in which the hip joint is replaced by an artificial implant. The hip implant consists of two components, a cup that is placed in the acetabulum of the pelvis and a stem that is inserted into the femur. A ball at the proximal end of the stem fits in the acetabular cup. Modern cups are made of either polyethylene alone or of a metal shell (i.e. titanium alloy, cobalt chromium alloy, stainless steel, and polymer composites) with a polyethylene liner. Stems are made of metal (i.e. titanium alloy, cobalt chromium alloy, stainless steel, and polymer compo-sites). Excellent functional results have been reported and today THA is considered a cost-effective and safe alternative to non-invasive treatment alternatives. Early postoperative mortality after elective primary THA is low and has decreased over the last years.6, 24, 35, 41, 47 THA has been so successful that it was even named “the operation of the century” in 2007.36

THA is used in the treatment not only of primary hip OA but also for de-formities, sequelae after childhood hip disease, rheumatoid arthritis, idio-pathic femoral head necrosis, secondary OA after trauma, tumours, and dis-located femoral neck fracture. However these patient groups vary in age, gender, comorbidities, and socioeconomic background. Over the years it has become evident that there is a need to take these differences into considera-tion when evaluating postoperative results.

18

THA- Different Fixation Methods There are different ways to fixate the chosen implants (Figure 1). Cemented implants are fixated using bone cement that chemically consists of polyme-thyl methacrylate (i.e. Plexiglas). Cementless implants are designed to allow bone ingrowth and the implant adheres over time. The term ”classic hybrid” fixation is used to describe the combination of a cementless implant in the acetabulum and a cemented implant in the femur. A ”reversed hybrid” is made up of a cemented implant in the acetabulum and a cementless implant in the femur. ”Resurfacing” describes an implant design that allows less of the original bone to be removed. The head of the femur is preserved and fitted with a metal cap dimensioned to fit into the acetabulum cup. The in-ternal diameter of the acetabulum cup is wider.

Figure 1: Cemented (left) and cementless (right) fixation of THA implants. © Åke Karlbom

19

It has previously been suggested that cemented fixation of THA increases perioperative mortality when compared with cementless THA, due to fat embolism and cardiopulmonary adverse events caused by the cementation process.41 However, in the cited work41 there was an evident issue of selec-tion bias where younger and healthier patients are selected for hip-resurfacing and hence that group has a lower mortality. After this study a debate followed regarding the importance of residual confounding and its role in the interpretation of register-based research results.31, 71

Evaluating Mortality After Total Hip Arthroplasty The most common outcome used for analyses of mortality in relation to a surgical intervention is 90-day mortality, since this parameter captures both perioperative deaths, early postoperative and delayed postoperative deaths such as those caused by thromboembolic events.46, 66 Hence, 90-day mortali-ty is the primary outcome in our studies. Two recent systematic reviews conclude that the average 90-day mortality after primary THA performed for any indication is 0.65 or 0.7%.9, 58

Other commonly used outcomes are 30-day mortality, 180-day mortality, and in-hospital mortality. 30-day mortality captures perioperative and early postoperative deaths - like 90-day mortality - but the capture of deaths caused by thromboembolic events and severe infections (i.e. pneumonia or urosepsis) is incomplete since the debut of symptoms related to these disor-ders often is delayed. 180-day mortality is more complete in capturing the delayed deaths related to the surgical procedure, but it also includes other deaths unlikely to be related to the surgery and is therefore more imprecise. In-hospital mortality does not capture deaths outside the hospital, as indicat-ed by the chosen expression, even though they may be related to the proce-dure. Taken together, the various measures of early post-operative death described above are all a trade-off between sensitivity and specificity when it comes to differentiating death related to surgery, and death due to other rea-sons.

Mortality After Bilateral Total Hip Arthroplasty Simultaneous bilateral surgery (i.e. one-stage surgery with both hips operat-ed on sequentially during one anaesthesia) is also described as feasible and safe, and early postoperative mortality after this procedure is reported to be low. A 90-day mortality of 0 to 0.14% 46, 66 for patients after simultaneous bilateral surgery and a peri-operative mortality (without specification of the exact time frame) of 0.31% (vs. 0.18% when compared with unilateral sur-gery) 32 are reported. There remain, however, concerns about the safety of

20

simultaneous bilateral THA procedures. Earlier studies have reported a high-er instance of haemorrhage, an increased number of thromboembolic events and cardiopulmonary complications, and a reduced range of motion after simultaneous bilateral operations 8, 39, 54.

Mortality After Total Hip Arthroplasty Due To Femoral Neck Fracture The mortality after THA due to hip fractures in the elderly is high compared to the elective THA procedures performed due to chronic diseases such as OA. In this patient group the mortality is around one per cent within the first 48 hours after surgery, if patients treated with internal fixation are included. 22 In the fracture group, early mortality after THA seems to be lower when compared to hemiarthroplasty or internal fixation. 30-day mortality has been estimated at 1.4% and 90-day mortality at 3.2% after THA due to hip frac-ture.28, 63 In comparison, an in-hospital mortality of 5.6% after hemiarthro-plasty and of 3.9% after internal fixation has been reported.64 However, this difference in mortality can presumably be explained by selection bias, i.e. less frail patients are selected for THA. One-year mortality after treatment of hip fractures with internal fixation or hip arthroplasties has been estimated at approximately 20%. There has been no obvious change in one-year mortality rates over the last three decades.42

Comorbidity Medical comorbidity is known to strongly influence the outcome of different medical interventions, including THA9, 24 (Figure 2), and there are different ways to investigate this influence. Several indices have been constructed for this kinds of analyses. The most commonly used would be the modified Charlson (Charlson/Deyo) comorbidity index (CCI)12, 15, the Elixhauser comorbidity score16 and the RxRiskV.17, 60 However, none of these were constructed for the purpose of investigating outcomes after THA.

21

Figure 2: Comorbidity. © The Swedish Hip Arthroplasty Register

Both the CCI and the Elixhauser score are based on ICD-codes. ICD is a classification system used since 1964 (ICD-7) in which every disease, condi-tion, and trauma event is given an alphanumeric code. This number can be used for epidemiological, health management, and clinical purposes.23 CCI can be calculated by using ICD-codes registered for outpatient visits, for in-hospital care or by using ICD- codes from both outpatient visits and from in-hospital care. However, outpatient visits were not included in the NPR until 2001, and hence our dataset is incomplete with regard to this parameter.

The Charlson Comorbidity Index is a diagnose-based coding algorithm introduced in 1987.12 It was developed through defining numerous clinical conditions and assessing their relevance in the prediction of one-year mortal-ity. In the original index, each of the 19 categories was assigned a weighted score based on their relevance in predicting one-year mortality. The original index was based on a breast-cancer reference population. The index has evolved over the years with modifications to the weights and to the list of specific International Classification of Diseases (ICD)-codes that are used.15,

48, 55 The most recent update of the CCI was made in 2010 by Quan et al who re-evaluated the index, reassigned weights to each condition and applied the index to populations from six different countries.49 For clarity I reasoned that only one CCI version should be included in studies II-III and chose the ver-sion modified by Deyo15, with the original weighting adapted to ICD-codes, since it is validated for the majority of the study period of 1992 to 2012. In study IV both the original weighting (Deyo) and the weighting according to

22

Quan were investigated.12, 49 In order not to underestimate the CCI calculated on ICD- codes, we used ICD-codes registered both for outpatient visits and for in-hospital care was included in our analyses (II and III).

The Elixhauser Score is a diagnose-based coding algorithm that was in-troduced in 1998.16 The Elixhauser Score is more detailed than the CCI, with 30 categories included in the original algorithm. The Elixhauser score has also evolved over the years. In study IV the weighting algorithm according to van Walraven was used. The van Walraven weighting algorithm is based on the association between comorbidity and death and gives an overall score for the Elixhauser Index.69

The RxRiskV Score is a prescription-based coding algorithm developed for research purposes and introduced in 2003.17, 60 In recent years, prescrip-tion based comorbidity measures have been increasingly used. It has been argued that a prescription-based algorithm would be more reliable than a diagnosed based algorithm, giving that it would have less of the limitations of incomplete or inaccurate coding.27, 29

CCI, the Elixhauser and the RxRiskV coding algorithms all have their limi-tations. They are based on the availability of extensive databases including a large number of ICD-codes or detailed information on drug prescriptions prior to surgery and are hampered by the usual limitations such as incom-pleteness and inaccuracy of coding.10 As described above, modifications of the weighting of certain items contained in the comorbidity measures and revalidations of the scores have been necessary over time.14, 15, 49, 65, 69 As a result, numerous variations of the same score co-exist, which can result in interpretation uncertainty. CCI, the Elixhauser and the RxRiskV are not used in clinical settings since the administrative burden is unrealistic. Thus, comorbidity measures based on patient administrative data are only accessi-ble to researchers. The clinician seeks a more easily accessible tool, i.e. the American Society of Anaesthesiologists physical status classification.

The American Society of Anaesthesiologists physical status classifica-tion (ASA) is a six-category physical status evaluation system developed in 1941.56 The ASA has remained unchanged except for a sixth category having been added (brain-dead person). Information on the ASA grade was included in the SHAR in 2008. The ASA grade is easily assessed in a clinical setting, but it is known to have a high degree of internal variability20 and it has not been validated as a predictor of mortality after THA. The ASA grade has previously been compared to the CCI, but not with respect to mortality after THA.33, 72

23

Table 1. The American Society of Anesthesiologists (ASA) classification

ASA Physical Status

1 Healthy person

2 Mild systemic disease

3 Severe systemic disease

4 Severe systemic disease that is a constant threat to life

5 A moribund person who is not expected to survive without the operation

6 A declared brain-dead person whose organs are being removed for donor purposes

In search of an alternative to the ASA grade, attempts have been made to quantify the individual patients’ risk profile. This has resulted in the creation of several universal and arthroplasty-specific risk prediction tools of varying degrees of complexity. However, none has been broadly accepted.40 Thus, an easily applicable tool with few dimensions is needed, both in research and in clinical practice.

Socioeconomic Background We know that socioeconomic background influences health and mortality.44,

74 Socioeconomic factors are also known to influence the outcome after THA.11, 18, 50 In order not to overlook the effect socioeconomic background has on the investigated outcome, it is advisable to adjust for it in a chosen statistical model. The variables most often used are educational level, in-come, and residence. Still, adjustment for socioeconomic background is lacking in many studies on mortality after THA. In Sweden, socioeconomic information can be collected from Statistics Sweden (see below). In 2004 Statistics Sweden changed the way by which the registered income variables are calculated. These changes unfortunately resulted in the variables being uncomparable before and after 2004. To overcome this problem I used quar-tile splits (II and III).

Sources of Epidemiological Data The Swedish Hip Arthroplasty Register The SHAR is a nationwide prospective THA database founded in 1979. The data was initially aggregated based on hospital units, but since 1992 the data are individualized based on personal identity numbers. The SHAR includes patient-related information (i.e. personal identity numbers, gender, diagno-sis, height and weight) as well as procedure-related information (i.e. date of

24

surgery, type of procedure, laterality, hospital, surgical approach and implant details). Since 2002 patient-related outcome such as pain relief, satisfaction and gain in quality of life have been included. Through linkage to the patient registers of the National Board of Health and Welfare, further analyses re-garding adverse events have been made possible. In 2014 the SHAR report-ed 100% coverage, with all hospitals performing THA participating and with 98% of all THAs reported.1 The coverage of the SHAR has been stable for many years. The SHAR has been repeatedly validated, and on a national level the completeness of registration has been stable at around 96-98% 2, 61,

62, although some studies have reported lower completeness for the endpoint revision surgery.2, 43

Statistics Sweden Statistics Sweden is a state-owned registry collecting information on the entire Swedish population. It includes detailed baseline demographics, i.e. level of education, income (both on a household and on an individual level), ethnicity and residence.

The National Patient Register The National Board of Health and Welfare is a government agency under the Ministry of Health and Social Affairs in Sweden. It has been collecting in-formation on in-patients treated in public hospitals in the National Patient Register (NPR) since the 1960s. In 1987 information from private hospitals was included, and since 2001 outpatient visits have also been included in the NPR. The registry has been validated on several occasions and has good external validity.34, 53, 68

The National Board of Health and Welfare has several other national reg-isters concerning the health of the Swedish population, i.e. the Medical Birth Register, the Cancer Register, the Drug register, and the Cause of Death Register.

The Drug Register The Drug register was started in 2005 by The National Board of Health and Welfare and includes information on all collected prescriptions. Information on prescriptions needed to calculate the RxRiskV score in study IV was col-lected from the Drug register.

25

Ethics and Legislation I feel it appropriate to reflect upon the ethics and legislation surrounding my research. Sweden gives an ideal platform for national observational studies with a long tradition of medical and socioeconomic registers, and personal identity numbers that enable linkage of data. Traditionally, the Swedish peo-ple’s trust in the government and the research community has been robust, which has facilitated the construction of broad national registers.

The ethical principles regarding human research are compiled in the Dec-laration of Helsinki.73 It was developed in 1964 by the World Medical Asso-ciation and the last revision was made 2013. It is not a legally binding in-strument but it has had a vast influence on legislation and regulations worldwide. The research subjects’ interests when it comes to integrity and autonomy are regulated as is the researchers’ obligation to ”do no harm” and to take responsibility for potential complications. These values are ever rele-vant and essential to protect.

In Sweden, research databases linked to personal identity numbers are on-ly allowed for a predetermined period of time in order to protect the research subjects’ integrity. You do not need to collect individual written consent for observational register studies. In order to protect the research subjects' au-tonomy while still enabling large observational studies, all research subjects are informed about being included in a register and are given the choice not to be part of the register or associated research at any time. This is regulated in the Swedish Patient Data Law of 2009 and the Personal Data Act of 1998.

Authorship of published scientific results is also protected by ethics and regulations. According to the Vancouver protocol25, “Each author should have participated sufficiently in the work to take public responsibility for the content.” In the more recent ICMJE guidelines26 it is stated that authorship also requires “Agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.” And yet, author lists sometimes have a tendency to lengthen unproportionally. I believe there still may be room for improvement.

Trust needs to be continuously earned. In order to keep society’s trust the science community needs to strictly adhere to our ethical principles.

THA is overall a safe procedure, but adverse events do occur. The most dramatic complication is death in the postoperative phase. Earlier research on the subject has shown that mortality after THA was low (90 mortality between 0.4%-0.8%) and that it had decreased over the last years.6, 9, 24, 35, 41,

47, 59 However, many previous studies had small study populations (i.e. n=7,4786, n=12,72859 and n=71,81235), and comorbidity and socioeconomic background were not included in most studies (for exceptions see Hunt24, Pedersen47, Whitehouse71). It remained unclear whether cemented fixation or

26

simultaneous bilateral THA surgery was associated with a higher early post-operative mortality. Due to the described diversity in the choice and size of study populations, and in the choice of included variables, the influence of comorbidity and socioeconomic factors on mortality after THA was also still unclear. In order to better understand how patient-related variables influence the outcome of THA surgery they need to be included in our analyses.

27

Aims

The overall aim of this thesis is to describe and investigate early mortality after THA. The specific study objectives are the following:

I To describe and investigate the risk of mortality when con-ducting a simultaneous bilateral THA compared with staged bilateral THA.

II To investigate early mortality and morbidity after THA per-formed in elderly patients with femoral neck fractures, adjust-ed for medical comorbidities and socioeconomic background.

III To investigate whether cemented fixation of THA increases early postoperative mortality when compared with cementless THA, with adjustment for medical comorbidities and socioec-onomic background.

IV To investigate the capacity of the following to predict early mortality after THA: a) established comorbidity coding algorithms i.e. the Elix-

hauser score, Charlson Comorbidity Index (CCI), and the prescription-based RxRiskV

b) easily accessible data that are routinely collected in daily clinical practice.

28

Methods and Study Populations

Ethics Ethical approval was applied for in 2013 and approved by The Regional Ethical Review Board in Gothenburg (2013: 360-13). All individuals regis-tered in the SHAR have received written information about the register and have access to the SHAR’s open webpage.

Methods The works of this thesis are all register-based. As mentioned in the introduc-tion, Sweden gives us an ideal platform for national observational studies thanks to the personal identity numbers. Every person who remains in Swe-den more than a year receives a personal identity number and, since 1947, is entered into the Total Population Register.38 The number consists of ten dig-its, based upon the date of birth and four added numbers where the last is a calculated control digit. Each personal identity number is unique for one resident in Sweden at a specific time, but the numbers are recycled. The identity number may show up several times in large register analyses over long time periods. These duplicates were excluded from the analyses since they represent different individuals.

Socioeconomic information was collected from Statistics Sweden (see above). Residence, level of education and personal as well as family income were registered per the year-end prior to index surgery.

Study I Study I was an observational register study with data obtained only from the SHAR. 42,238 patients that had received bilateral primary THA between 1992 and 2012 and that had been reported to the SHAR were included in the analyses. For comparative reasons patients who had had a unilateral primary THA during the same time period (1992 to 2012) were also analysed. This “unilateral cohort” consisted of 148,718 patients.

Main outcome was 90-day mortality after simultaneous bilateral THA compared with staged bilateral THA. Secondary outcomes were

29

30-day-, 10-year-, and overall mortality after the second THA surgery. Revision surgery of any hip after the second THA surgery was also explored as an additional outcome.

Study II Study II was designed as a nation-wide matched cohort study. 24,699 pa-tients in the SHAR who had received a THA due to a femoral neck fracture were compared with 118,518 controls from the general population. Patients with pathological fractures on the femoral neck were excluded. In patients with bilateral THA, we included only the side which was operated on first. The controls were matched for age, gender and residence, and in multivaria-ble analyses adjustment was made for medical comorbidity and socioeco-nomic background. Primary outcome was death for any reason within 90 days of the index surgery. Secondary outcomes included 90-day mortality for cardiovascular reasons and the risk of admission to hospital care within 90 days from the index date.

Study III Study III was designed as a nation-wide matched cohort study of primary THA patients and matched controls, with adjustment for medical comorbidi-ty and socioeconomic background in multivariable analyses (Figure 3). 178,784 primary THA patients with primary and secondary OA and 862,294 controls matched for age, gender and residence were included. In bilaterally operated patients, only the first surgery was included. Patients with femoral neck fractures, rheumatoid arthritis, and tumours were excluded. The main outcome was mortality within the first 90 days after surgery. The unadjusted survival curves of the THA patients and the controls crossed around day 15. The observation period was therefore divided into the three periods: day 0 to 14, day 15 to 29, and day 30 to 90.

30

Figure 3: Flowchart for study III.

Study IV Study IV was designed as a nationwide prospective observational cohort study. 38,735 patients who had received a THA due to primary OA between 2008 and 2012 were included. The starting point of 2008 was chosen be-cause ASA grade and BMI were included in the SHAR from that year on-wards. In order to avoid dependency issues, only the first surgery was ac-counted for in bilaterally operated patients. Patients were excluded if a se-cond THA surgery was performed within 12 months after the first THA sur-gery. Clinically easily accessible data, i.e. ASA grade, BMI, age, and gender, were investigated, as were three commonly used comorbidity coding algo-rithms i.e. CCI, Elixhauser, and RxRisk V, with their respective included dimensions. Relevant candidate predictors from the separate dimensions of the comorbidity algorithms were identified based on lower bounds of the CI being larger than one and were combined in a larger model together with age, gender, BMI, and ASA. Predictors were excluded from the model one at a time until the combination of predictors with the highest degree of predict-ability was found. Primary outcome measure was 90-day mortality after THA surgery. Both the risk of death and the degree of predictability for po-tential predictors separately and in combination was analysed.

31

Statistical Methods In the linkage process and building of the original combined database for study II and III, assistance was provided from statisticians at the SHAR, Statistics Sweden, the National Board of Health and Welfare, and from Upp-sala Clinical Research Centre. Information from the different registers was linked by personal identity number. 5,572 individuals had recycled or incor-rect identity numbers and hence were excluded. If the excluded individual was a case, then the respective controls were also excluded. Some controls had emigrated or died prior to index date, as controls were age-matched by year but not by exact date of birth. These controls were excluded. Surpris-ingly, 11 controls also had information in the SHAR. These were also ex-cluded.

For all four studies, follow-up started on the day of the THA surgery and ended on the day of death, emigration, or December 31st 2012, whichever came first. Some death dates were unspecific: “Month X 13th 2001”, “Month X Yth 2001”, or “August Yth 2001”. Where this occurred, the unknown month was set to “May” and unknown day to “15” in order to avoid over/under estimating the survival time.

For comparative reasons I chose to primarily use the CCI in studies II and III. CCI has been calculated separately for outpatient visits, hospital admis-sions, and for all healthcare contacts together. The modified CCI was calcu-lated based on the ICD-codes registered one year (366 days) prior to the index date for each individual. In the described analyses on databases reach-ing back to 1992, the ICD 9 and 10 classification systems are used. The ICD 9 and 10 codes used are identical to those described in Quan’s adaptation of the Charlson/Deyo algorithm.15, 48, 65 ASA grade was first included in the Swedish Hip Arthroplasty Register (SHAR) in 2008 and was therefore not available for all cases. It was by definition unavailable for all controls since it is only registered in the medical records prior to surgery and reported to the SHAR but not to the NPR.

The choices of covariates included in the different multivariable regres-sion models were based upon assessment of relevance and non-interference using directed acyclic graphs as previously suggested57 (example Figure 4). I have adhered to the suggested guidelines on statistical analyses of register data51, 52.

32

Figure 4: Directed acyclic graph for study I.

Mortality within 90 days after THA is a rare event. Assuming that 90-day mortality is around 0.7 %, 1,346 exposed individuals and 13,460 controls would be required to detect a difference of factor 2 according to the per-formed power calculations (alpha=0.05, power 80%). For comparison, if a difference of factor 1.5 is to be detected, 5,831 exposed individuals and 29,153 controls would be required (alpha=0.05, power 80%).

Continuous data were described using means, medians and ranges, and 95% confidence intervals (CI) were used to describe estimation uncertainty. Categorical data were cross-tabulated and a Chi-square test was used to in-vestigate whether observed and expected frequencies differed significantly between groups of data. The chosen age categories (< 50 years, 50-59, 60-74, and ≥ 75 years) are the same age categories that are used in the annual reports of the SHAR.

Unadjusted cumulative survival was calculated using the Kaplan-Meier method.

In studies I to III, Cox regression models were fitted for each covariate at a time in order to calculate crude hazard ratios (HR) with CI. Covariates were then subsequently included in multivariable regression models in order to calculate adjusted HR with CI (Adjustment for gender, age, Charlson’s comorbidity index and socioeconomic background data i.e. personal income and education). Model assumptions were investigated by calculating and plotting the correlation coefficient between transformed survival time and the scaled Schoenfeld residuals. No imputation for missing data was per-formed.

33

The level of significance was set at p<0.05 in all analyses. The R software (package 3.0.232) and SPSS (version 22) were used.

In studies II to IV, only the first surgery was accounted for in bilaterally operated patients in order to avoid dependency issues, i.e. the number of investigated procedures and the number of THA patients is identical.

Given that neither age nor income are treated as linear variables, age and income in study III were additionally modelled by restricted cubic splines, also known as natural splines, in separate analyses. In order not to over-fit the spline, the number of knots was chosen by the Bayesian information criterion. Theoretically it is possible to increase the likelihood of an event occurring by adding parameters to a chosen statistical model, but this may result in overfitting. The Bayesian information criterion attempts to resolve this problem by introducing a penalty term for the number of parameters in the model.

In studies II and III, results on relative mortality risks derived from Cox regression models were also validated by fitting Aalen additive risk models,4 which have been suggested to have advantages over Cox regression models. The main advantage of the additive risk model is that no assumption of pro-portionality needs to be fulfilled, since no time variable is included in the calculation. In that sense, an Aalen additive risk model is rather similar to a logistic regression model.

In study IV the expected number of events to be censored was relatively low. I therefore chose logistic regression models and not Cox regression models as in studies I - III. These logistic regression models were fitted for each covariate at a time in order to calculate crude hazard ratios (HR) with CI. Subsequently ASA, Elixhauser, CCI, and RxRiskV were one at a time combined with age, gender, and BMI in four separate models. The perfor-mance of the logistic regression models was evaluated by its discrimination capacity using c-statistics. The c-statistic represents the area under the curve (AUC in the plotted ROC [receiver operated characteristic] graph) and stands for the degree of predictability. An AUC of 50% (0.5) implies a dis-crimination capacity equal to pure chance (i.e. if A there is a 50% chance of B). An AUC of 100% (1.0) implies a definite predictive relation (i.e. if A then B in 100% of cases). All relevant candidate predictors were combined in a larger model together with age, gender, BMI, and ASA. Through back-wards elimination, the predictors were excluded one at a time, until the op-timal combination with the highest degree of predictability was found. Then the excluded predictors were reinserted one at a time to see whether this would strengthen the degree of predictability with statistical significance. In study IV, I also chose to keep all continuous variables continuous in order not to weaken the precision of the statistical analyses.5 No imputation for missing data was performed.

34

Results

Study I During the study period patients selected for simultaneous bilateral surgery were younger, male, and had lower ASA grade than patients receiving staged procedures. The adjusted 90-day mortality after the second procedure did not differ between the four investigated groups. Patients older than 75 years at index surgery, men, patients with ASA grade 3 or above and patients with rheumatoid arthritis (RA) had an increased 90-day mortality (Table 2). The unadjusted risk of implant revision of any hip was slightly higher for patients with simultaneous bilateral THA compared to those with a staged procedure within one year, but after adjustment for age, gender, diagnosis and implant fixation these differences were no longer statistically significant (Table 3).

35

Table 2: Hazard ratio for 90-day mortality in study I. Adjustment for sex, age, diag-nosis, type of prosthesis fixation and time interval.

Table 3: Risk of implant revision in study I. Adjustment made for sex, age, diagno-sis and type of prosthesis fixation.

No of events

Crude HR (CI 95%)

Adjusted HR (CI 95%)

Simultaneous 240 (7.1%) 1.6 (1.3-1.9) 1.1 (0.9-1.4) ≤6 months 495 (5.1%) 1.2 (1.1-1.4) 1.1 (1.0-1.3) 7- 12 months 828 (5.3%) 1.1 (1.0-1.3) 1.1 (1.0-1.2) >12 months1 3 573 (6.4%) - -

1 Reference group

36

Study II Crude survival after femoral neck fracture for all THA patients together was lower than that of control individuals over the entire observation period. I found a 3.0-fold (CI 2.7 to 3.2) increase in the crude risk of death within the first 90 days after the index date for THA patients compared with control individuals, i.e. the 90-day survival after the index date was 96.3% (CI 96.0 to 96.5) for THA cases and 98.7% (CI 98.6 to 98.8) for control individuals.

Table 4: Crude and adjusted risk of 90-day mortality in study II with 95% confi-dence intervals. Adjusted hazard ratios (HR) derived from a model adjusting for gender, age, Charlson’s comorbidity index and socioeconomic background data (personal income and education). Reference groups by definition have a HR of one.

Risk of 90-day mortality

Crude HR

95% CI

Adj. HR

95% CI

Case1 Arthroplasty 2.96 2.72-3.21 2.18 2.00-2.37 Gender2 Female 0.69 0.63-0.75 0.71 0.65-0.78 Age 60-69 0.53 0.44-0.64 0.58 0.48-0.70 80-89 2.64 2.39-2.90 2.32 2.10-2.56 90-99 6.23 5.47-7.10 4.80 4.16-5.55 CCI3 1-2 4.03 3.69-4.41 3.14 2.87-3.44 >2 10.43 9.36-11.63 8.41 7.52-9.41 Income4 2nd quarter 1.04 0.94-1.15 1.02 0.92-1.13 3rd quarter 0.88 0.79-0.98 0.95 0.85-1.06 4th quarter 0.57 0.50-0.65 0.84 0.73-0.97 Education5 9 years 0.46 0.42-0.51 0.74 0.66-0.83 High School 0.31 0.27-0.35 0.57 0.49-0.66 University 0.24 0.20-0.29 0.50 0.41-0.61

1 Reference = Controls 2 Reference = Male 3 Reference = 0, CCI= Charlson Comorbidity Index 4 Reference = First quarter (lowest income) 5 Reference = No school education or unknown level of education

37

Adjusted 90-day mortality in femoral neck fracture patients receiving a THA was about double compared with a matched control population. The most pronounced risk increase was associated with the presence of a CCI of three or more. An age of 90 years or above was also associated with a high in-crease in the risk of 90-day mortality. Higher levels of education were asso-ciated with a risk reduction while income had smaller effects on the adjusted risk of 90-day mortality (Table 4). Patients below 80 years of age and with-out significant comorbidity had a mortality of about the same magnitude as controls.

The findings in studies I, II and III were derived from multivariable Cox regression models. The validity of Cox regression models relies on the as-sumption of proportional hazards. A typical deviation from the assumptions of the Cox model is the potential time-varying effects of different covariates. However, there are alternative statistical models used to alleviate this prob-lem. In studies II and III Aalen's additive risk model was used as a comple-ment. In Aalen's additive risk model the effect of the covariates acts on an absolute rather than a relative scale. In study II, Aalen’s additive models confirmed the increase in the adjusted risk of early mortality for hip fracture patients treated with a THA compared to controls (Figure 5).

Figure 5. The additive mortality risk for THA fracture patients compared to control individuals in study II. Aalen’s additive model with 95% confidence intervals.

38

Study III In this study population there were more females than males, and the age at surgery varied between 15 and 100 years with a mean of 68.9 years and a majority in the age category 60-74 years. The characteristics of the two sub-cohorts based on fixation method are presented in Table 5. Since THA cases and controls were matched for age and gender the distribution into the cho-sen gender and age categories was quite similar.

Table 5: Characteristics of the study population in study III, divided by cases (ce-mented and cementless THA) and controls.

An increased adjusted risk of death during the first 14 days was found in patients that received cemented THA compared to controls. This corre-sponds to an absolute risk increase of five deaths per 10,000 observations. From day 15 to 29 the mortality risk was lower for patients receiving ce-mented THA. No early risk increase was observed in patients receiving ce-mentless THA. This finding was confirmed with Aalen’s additive risk model (Figure 6).

The unadjusted survival was lower for those with a classic hybrid fixation compared to their controls. This was not observed in patients who received a reversed hybrid. The same pattern was seen after adjustment for age, gender, comorbidity, and socioeconomic background: an increased risk of mortality during the first 14 days for the classic hybrid THA compared to controls.

39

This effect could not be found among reversed hybrid THA compared to their controls.

Figure 6: The additive mortality risk for cemented and cementless THA compared to controls in study III. Aalen’s additive model with 95 % confidence intervals. Time in years.

40

Study IV This study evaluated to which extent the easily accessible data that are rou-tinely collected in daily clinical practice could be used to substitute a more complex comorbidity score for prediction of 90-day mortality after THA. In the study population there were more females (56.2%) than males (43.8%), and the age at surgery varied between 18 and 100 years with a mean of 68.9 years. The characteristics of the study population in study IV are very similar to those of study III. This is expected since the cohort of study IV is also a part of study III’s study population.

Unadjusted cumulative 90-day survival was 99.7 (CI 99.68 to 99.78) and the number of deaths during this period was 109. As in studies I to III, I found that age (OR 1.1 [CI 1.09-1.14]) and male gender were associated with an increased risk of death (OR for female gender: 0.5 [CI 0.37-0.79]). Patients with a higher degree of comorbidity according to ASA, the CCI, the Elixhauser score and the RxRiskV score had an increased risk of 90-day mortality compared with patients with a lower score.

I found that the combination of age, gender, history of heart infarction and renal disease, and ASA grade was the best predictor of both 90-day mor-tality (c=0.80) and one-year mortality (Figures 7 and 8). Introducing socio-economic variables such as education level and marital status into the multi-variable regression model only marginally increased its predictive power. Adding information on BMI to the model did not strengthen the predictive power. The ASA classification alone had a better predictive strength for 90-day mortality than the more complex comorbidity measures CCI, Elixhauser score and RxRiskV score (Figure 7).

41

Figure 7: ROC for the investigated comorbidity measures in study IV: New model with age, gender, ASA, history of heart infarction and renal disease during the last 12 months included, ASA alone, CCI, Elixhauser and RxRiskV.

Figure 8: ROC for 90-day and one-year mortality for age, gender, ASA, history of heart infarction and renal disease during the last 12 months combined from study IV.

42

Further Results Comorbidity- Trend over time By choosing to include the CCI based on ICD-codes registered both for out-patient visits and in-hospital care, I was able to capture more comorbidities than I would have otherwise (Table 6). THA patients had more comorbidi-ties than controls, even when only taking outpatient visits into account (Ta-ble 6).

Table 6: CCI in the original study population of study III. Cemented and cementless THAs and controls. Only cases with complete data included. Outpatient visits were not included in the NPR until 2001 and are accounted for as CCI=0 before then.

Outpatient Care

In Hospital Care

Outpatient and In Hospital Care combined

CCI 0 (%) Arthroplasty 92.2 86.8 82.8 Control 93.2 92.4 87.9 CCI 1-2 (%) Arthroplasty 7.0 12.1 15.2 Control 5.9 6.3 10.0 CCI >2 (%) Arthroplasty 0.8 1.1 2.0 Control 0.9 1.3 2.1 Mean CCI1 Arthroplasty 0.14 (0-11) 0.20 (0-13) 0.28 (0-14) Control 0.13 (0-12) 0.15 (0-14) 0.23 (0-15)

1 Mean CCI and ranges.

43

Over time a slight increase in the presence of any comorbidity was seen in the study population of study III. For the period 1992-1996 the combined outpatient and in-hospital CCI was 0 for 87.0% of all observations, 1-2 for 10.9% and >2 for 2.1%. During the period 2007-2012 the combined outpa-tient and in-hospital CCI was 0 for 82.9%, 1-2 for 13.9% and >2 for 3.2%. The same trend was seen when looking at THA patients and controls sepa-rately.

Mortality- Trend over time In study III a decrease in mortality over time was seen. The numbers shown in Table 7 are for the study population of cemented and cementless THA.

Table 7: Trend over time in crude survival in study III with 95% confidence inter-vals.

Crude 90-day survival

% CI (95%) 1992-1996 Arthroplasty 99.5 99.44-99.51

Control 99.2 99.22-99.26 1997-2001

Arthroplasty 99.4 99.31-99.48

Control 99.2 99.14-99.22

2002-2006

Arthroplasty 99.6 99.59-99.70

Control 99.3 99.22-99.30

2007-2012

Arthroplasty 99.7 99.61-99.71

Control 99.4 99.33-99.39

44

Discussion

There is still much to be done to increase our knowledge of mortality after THA surgery. There are too many studies in the literature with study popula-tions that are too small to achieve sufficient statistical power, that lack a relevant control group, or that are insufficiently adjusted and thus do not account for socioeconomic background and comorbidity. Early postoperative mortality after THA is so low that it is difficult to gather many patients and events in one single hospital or region. Studies on nationwide registries can therefore be of great help. Sweden, with its system of personal identity num-bers and its national registries, is an ideal platform for analyses of mortality after THA.

Based on the results in study I, I concluded that simultaneous bilateral THA can be considered a safe alternative in Sweden. I was not able to adjust for comorbidity or socioeconomic background in this study, however. For this reason, caution should be taken in patients with advanced age, male gender, rheumatism and high ASA grade since these variables were associat-ed with an increased risk of 90-day mortality.

In study II I found that hip fracture patients treated with THA had a more than two-fold adjusted mortality rate within the first 90 days after the index date compared with a matched control population. High mortality in this patient group has previously and repeatedly been documented, but not in such a well-defined group (i.e. all THA patients in a country) and after ad-justment for comorbidity and socioeconomic factors. Our findings indicate that THA surgery for femoral neck fracture is a comparatively safe alterna-tive in patients under the age of 80 and with few or no comorbidities. In more frail and older patients, however, alternatives to THA could be consid-ered. THA patients in our study were considerably sicker with a higher CCI at the time of injury than their matched controls, which is in line with previ-ous findings on patients with hip fractures. The high degree of comorbidities in these patients means that they are especially susceptible to falls resulting in fractures.

Both early postoperative and long-term mortality after primary THA due to OA was reduced in THA patients compared to controls. This is consistent with previous studies.7, 21, 37, 45, 47, 70 The prevailing explanation for this phe-nomenon is that patients selected for THA are healthier than the general population. This assumption seems to be wrong, at least for our study popu-lation: in study III I found that control individuals were healthier than pa-

45

tients selected for THA, at least when measuring comorbidity along the defi-nitions suggested in the CCI. Of course, a certain amount of detection bias could explain this phenomenon, i.e., an increased number of healthcare con-tacts for patients scheduled for THA prior to surgery could give rise to comorbidities being detected which would otherwise have remained obscure (Figure 9).9, 35 When calculating the CCI, only which diagnoses are present is relevant, not how many times the diagnoses have been registered. Chronic illnesses relevant for calculating the CCI should have been registered be-forehand and any emergency illnesses relevant for calculating the CCI ought to have been an absolute contraindication for surgery. It is doubtful that a sufficient number of relevant new diagnoses would be registered in connec-tion to surgery to explain the difference in CCI between THA patients and controls.

Figure 9: © The Swedish Hip Arthroplasty Register

Comorbidity is not static over time. Using the SHAR, Gordon et al. have previously demonstrated an increase in the presence of any comorbidity over time on a cohort of patients that received primary THA due to OA.19 The same phenomenon could be seen in the population investigated in study III, which is not surprising considering it includes some of the same patients present in Gordon et al.’s study. Despite an increasing comorbidity over the study period, we observed a decreasing early postoperative mortality - a finding consistent with earlier studies.6, 24, 35, 41, 47 Both crude and adjusted mortality decreased over time, which is reassuring for patients undergoing

46

THA. The decreasing early postoperative mortality could perhaps be ex-plained by better pre- and post-operative care, and the increasing comorbidi-ty by a more consistent registration of comorbidity diagnoses.

Socioeconomic background variables such as residence, educational level, income, and marital status are known to influence the availability and out-come of medical intervention44, 71, 74, i.e. THA.11, 18, 50 Still, they are often not adjusted for in analyses of morbidity and mortality after THA. In studies II to IV I found a distinct influence of the included socioeconomic factors on early mortality. However, in study IV level of education and marital status only marginally increased the AUC of the statistical model that attempted to predict early mortality after THA.

I believe there is a consensus that comorbidity has an influence on the outcome after THA surgery, but there remains an uncertainty as to how to measure comorbidity. CCI and Elixhauser are two of the most commonly used comorbidity measures in register-based research. According to our re-sults from study IV, however, a far less data demanding comorbidity meas-ure - the combination of age, gender, history of heart infarction and renal disease and ASA grade - serves us better in predicting early postoperative mortality after THA. Even ASA grade alone had a better predictive strength than comorbidity measures based on more complex coding algorithms. Ad-vancements in information technology offer us impressive possibilities to combine large quantities of data in our research analyses. But the proverb “Just because we can, doesn’t mean we should” still stands. I see the result of study IV as an indication that we may go back to clinical basics and that it may be advantageous to let those clinical variables be heard/analysed in large observational data based research as well.

47

Methodological Considerations When reviewing the literature for studies on mortality after THA, one quick-ly finds discrepancies in the choices of reference groups (Table 8).

Table 8: Chosen control groups in studies on mortality after elective primary THA due to OA. N= number of studies, SMR= Standardized Mortality Rate.

Choice of Reference N Authors

SMR 3 Lie SA et al,Parry M et al, Mahomed NN et al

Selected group within reach of the department

2 Jones D et al, Visuri T et al

Nationwide cohort 1 Pedersen AB et al

No unoperated control group

3 Mc Minn et al, Hunt LP et al, Whitehouse et al

A control group that is comparable to THA patients with respect to gender, age and socioeconomic background is desirable for correct interpretation of data, but such a control group has only rarely been available in previous studies30. In studies II and III, I attempted to relate mortality and morbidity in the population of THA patients to an adequate control group taken from the general population in order to strengthen the clinical relevance of our data. The control group was matched for age, gender, and residence. In stud-ies II and III there is adjustment for socioeconomic background and comor-bidity. A limitation of study I is the lack thereof.

In study I the validity could be threatened by the fact that patients sched-uled for bilateral procedures but that suffer complications during or after surgery No 1 never undergo surgery No 2. That would result in an underes-timation of the mortality associated with staged bilateral arthroplasty.

All registry studies are based on observational data with possible biases at different levels.

Selection bias: As clinicians we tend to select certain patients for certain procedures. We are taking plausible influencing variables into account in our risk evaluation, i.e. age, gender, comorbidity, etc. We can never with certain-

48

ty fully compensate for the issue of selection bias with this kind of study design.

Detection bias: In study III, I found that control individuals had fewer comorbidities than patients selected for THA. As mentioned above, this could be an example of detection bias where, prior to surgery, an increased number of healthcare contacts for patients scheduled for THA could give rise to comorbidities being detected that would otherwise have remained ob-scure.

Performance bias: This is the presence of systematic differences between groups in the care that is provided, i.e. healthcare routines that differ geo-graphically, between hospitals and even between individual surgeons (Figure 10). In studies II and III THA patients and controls were matched by munic-ipality, but this does not fully compensate for the performance bias issue.

Figure 10: © The Swedish Hip Arthroplasty Register

Immortal time bias: This occurs when individuals in one of two groups that are compared are guaranteed to be alive for period of time during the follow up time. In studies II and III both THA patients and controls are alive at the date of exposure, but in study I this phenomenon has not been accounted for.

Confounders: In all research, confounders can be out of reach or not cor-rectly adjusted for. The residual confounding in studies II - IV ought to be considerably smaller than in many previous studies since socioeconomic factors were included in our analyses.41, 47 Nevertheless, the possibility of

49

residual confounding remains due to other variables that are unknown and unaccounted for.

Validity: The registries used for our analyses are nationwide and have been repeatedly validated.2, 34, 53, 61, 62, 68 However, the population-based and nationwide selection of patients in studies of the kind presented in this thesis gives a high external validity (the results can be generalized to other set-tings) but a lower internal validity (the differences in outcome actually de-rive from the investigated exposure). The results can be generalized to other countries with similar socioeconomic and healthcare conditions. The na-tionwide design also provides us with enough statistical power to detect rare events. Mortality after THA is low, and hence a large number of patients are needed in order to detect a difference in mortality between different groups (i.e. cemented/cementless, male/female, different educational levels). Never-theless, there is the risk of detection bias mentioned above. Let us take comorbidity as an example: During the study period 1992-2012 there were changes in the system of reimbursement in the Swedish healthcare system. More registered diagnoses mean more money for the healthcare provider, to put it rather crudely. Perhaps the higher comorbidity in the later part of the study period is a sign of better registration routines due to financial incen-tives rather than less healthy patients being selected for surgery. That would be more consistent with the mortality decrease over time.

It is important to make a distinction between explanatory research and pre-dictive research. In the latter, predictive research, the investigated temporal context is another i.e. futurum. In order for a risk factor to be considered a predictor, the investigated effect needs to be tested in a different sample of individuals to capture the “futurum” aspect. In study IV, an external valida-tion has not been done. Hence it is only with extreme caution that we can extrapolate our findings in study IV to predictions in the clinical setting. Changes in comorbidities and mortality over time as observed in this thesis indicate that prediction models derived from data collected in the past might lose some of their prediction power for future surgical intervention, since the underlying data set will be different and, at least after some time, lose some of its relevance.

50

Conclusions

I With the present patient selection made in Sweden, simultaneous bilat-eral THA can be considered a safe alternative. However, caution should be taken in patients with advanced age, male gender, rheumatism and high ASA grade since these variables were associated with an increased risk of 90-day mortality.

II THA surgery for femoral neck fracture can be considered a safe alterna-tive for patients with few or no comorbidities under the age of 80. Other treatment options could be considered in older and/or more frail patients.

III The increase in absolute mortality risk up to 14 days after elective THA

surgery for OA using fully cemented implants is small and must be weighted against the previously reported increased risk of early revisions in connection with cementless implants. However, the patient’s socioec-onomic background and the presence of comorbidities have an important influence on the outcome and must be evaluated by the clinician.

IV Future mortality studies could benefit from the use of data that are rou-

tinely collected, and the logistically complicated procedure of merging national patient registries and databases might be avoided.

Decisions regarding which patients should be offered what treatment cannot be taken lightly. Primum non nocere.

51

Future Research

This thesis focuses on mortality after THA. However, mortality is only one adverse event that is of importance to the patient. Investigation of other ad-verse events, i.e. risk of infection or revision, including comorbidity and socioeconomic background, would also be of value.

In study I we found that revision surgery was slightly more common among those receiving simultaneous replacements compared to patients op-erated in a staged procedure. It would be interesting to analyze this further. I also find it tempting to investigate the simultaneous replacement group more closely, i.e. a journal review including causes of death.

In study II, patients who had THA surgery after hip fracture were com-pared with a matched controlled population. It would be interesting to repeat the study design for internal fixation and hemiarthroplasty surgery after hip fracture.

Study IV opens the door for future research in the area of using routine clinical data to predict mortality after THA. In order to extrapolate the find-ings from our prediction model (including gender, ASA, history of heart infarction and renal disease during the last 12 months) into a predictive tool in the clinical setting, the results would need to be validated in a separate cohort. Our study is a first step, indicating that it can be done. This valida-tion still needs to be done, preferably in cooperation with another national arthroplasty register in a country with similar socioeconomic and healthcare conditions.

52

Project Collaborators

Supervisors Nils Hailer, MD – Orthopaedic surgeon and professor at the Department of Orthopaedics, Uppsala University Hospital, Institute of Surgical Sciences, Uppsala University.

Olle Nilsson, MD, PhD – Orthopaedic surgeon and professor emeritus at the Department of Orthopaedics, Uppsala University Hospital, Institute of Sur-gical Sciences, Uppsala University.

Johan Kärrholm, MD, PhD – Orthopaedic surgeon, professor at the De-partment of Orthopaedics, Institute of Clinical Sciences, Sahlgrenska Acad-emy, University of Gothenberg and director of the Swedish Hip Arthroplasty Register.

Göran Garellick MD, PhD – Former director of the Swedish Hip Arthro-plasty Register and professor at the Department of Orthopaedics, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenberg.

Collaborators Ola Rolfson MD, PhD – Orthopaedic surgeon, associate professor at the Department of Orthopaedics, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenberg and director of the Swedish Hip Ar-throplasty Register.

Cecilia Rogmark MD, PhD – Orthopaedic surgeon, associate professor at the Department of Orthopaedics, Skåne University Hospital/Lund Univer-sity, and national project leader of the Swedish Hip Arthroplasty Register.

Max Gordon MD, PhD – Orthopaedic surgeon, researcher at the Depart-ment of Clinical Sciences, Danderyds Hospital, Karolinska institute.

53

Olof Sköldenberg MD, PhD – Orthopaedic surgeon, associate professor at the Department of Clinical Sciences, Danderyds Hospital, Karolinska insti-tute.

Szilárd Nemes PhD – Statistician at the Swedish Hip Arthroplasty Register.

Niclas Eriksson PhD – Statistician at Uppsala Clinical Research Center, Uppsala University.

Fereshte Ebrahim – Statistician at the National Board of Health and Wel-fare.

Andreas Blomquist – Statistician formerly at Statistics Sweden.

54

Acknowledgments

Nils Hailer You didn’t have any doubts about letting a tipsy dinner conversation end up in a serious thesis proposal. I was your first PhD student at a distance. I learnt a lot. You might have learnt some? I will miss you Nils, your endless enthusiasm, your deceivingly innocent questions pushing me further onto scientific ground, and your never-ending support. Thank you. For every-thing. Perhaps I will now start calling you “Nisse”.

Olle Nilsson Everyone I’ve met who has worked with you describes you as a role model. Thoughtful, kind and humble, your way of action at the clinic, in research and towards fellow human beings is truly inspiring. It has been an honor to strive the scientific path with you by my side. Thank you for believing in me and for being my friend.

Johan Kärrholm Thank you for your generous contribution to the substance of my work and to the direction of my scientific journey. You have always given me time, presence, and support, and several times you have been Nils’ stand in when I’ve needed it the most. You are not only the earliest early bird I have ever met but also one of the best chefs. Sometimes in the reverse time order though…

Göran Garellick You have re-introduced me to the very usable communication tool called “the phone”. I find it very effective but that can of course be that it has been you answering my questions at the other end. For those answers I thee thank, and for the champagne.

55

Li Felländer-Tsai If someone ever volunteers to be your academic mentor “from the outside” I would suggest that you accept. I did. Thank you Li, for volunteering to take me on, for ongoing discussions, inspiration, and tricks of the trade. I am now volunteering to keep seeing you even without this thesis project. I hope you accept. B&B.

Åke Karlbom Pater Noster Ortopaedica. I still haven’t met anyone as gifted as you with a scalpel and a saw. Thank you for introducing me to Orthopaedics, for being my clinical mentor and friend, and for your support in my undertaking the scientific journey you never undertook yourself. Thank you also for taking the time to illustrate my book. I look forward to spending more time with you at the clinic and I am very grateful that I have the possibility to do so.

Ola Rolfson For being an appreciated co-writer and for rapidly giving me a “How to sur-vive as a PhD-student” crash-course when I was about to start. And I know you know I needed that. Your generous often read e-mails full of tips and tricks have been very appreciated and they are now in the archive for safe-keeping. Thank you.

Hans Mallmin For having brought me to Uppsala, forgiven me for declining another thesis proposal, and for your interest in my work. Last but not least, thank you for teaching me that enjoyable opera music is heard not only at the Metropolitan and la Scala but also at the Royal Opera in Stockholm and at Göteborgsop-eran. I value your friendship.

Tönu Saartok, Carina Karlsson, fellow colleagues at the Department of Orthopaedics, Visby Hospital I know I haven’t been around as much as needed. Thank you for letting me, for not giving me a hard time about it, and for every now and then asking me how things are coming along.

56

The ultimate Grandmothers Eva & Ylva, Linda & Magnus, and Kerstin & Bosse You made it possible to sell waking up at 05.15 a.m. and having breakfast somewhere else/ in somebody else’s company as an adventure to a three year old. I don’t see how the logistics of this project, three wild ones, and on call duties for Martin and myself would have worked without you. Thank you.

“Bokis”, and Crossfit Visby For keeping me mentally and physically sound.

Pernilla & Samuel For your ability to remind me what is truly important in life.

Moa, Ingrid & Agnes You are truly important. Being the amazing, strong minded, kind, stubborn, helpful, cranky, curious, demanding, loving, wonderful kids that you are, you have helped to keep me balanced. It is a privilege to be your mother.

Martin If I could I would give you “Best Supportive Husband Award” retrospective-ly for the last 16 years in general, and for the last four years in particular. Thank you. You being you, enable me to be me.

57

References

1. Swedish Hip Arthroplasty Register (SHAR). The Annual Report 2014. http://www.shpr.se/. 2. Swedish Hip Arthroplasty Register Annual Report. 2012 http://www.shpr.se/sv/Publications.aspx. 2012. 3. R Development Core Team. R: A language and environment for statistical computing. 2012. http://www.R-project.org/. 2012. 4. Aalen OO. A linear regression model for the analysis of life times. Stat Med 1989; 8(8): 907-25. 5. Altman DG, Royston P. The cost of dichotomising continuous variables. Bmj 2006; 332(7549): 1080. 6. Aynardi M, Pulido L, Parvizi J, Sharkey PF, Rothman RH. Early mortality after modern total hip arthroplasty. Clinical orthopaedics and related research 2009; 467(1): 213-8. 7. Barrett J, Losina E, Baron JA, Mahomed NN, Wright J, Katz JN. Survival following total hip replacement. The Journal of bone and joint surgery American volume 2005; 87(9): 1965-71. 8. Berend ME, Ritter MA, Harty LD, et al. Simultaneous bilateral versus unilateral total hip arthroplasty an outcomes analysis. The Journal of arthroplasty 2005; 20(4): 421-6. 9. Berstock JR, Beswick AD, Lenguerrand E, Whitehouse MR, Blom AW. Mortality after total hip replacement surgery: A systematic review. Bone Joint Res 2014; 3(6): 175-82. 10. Bozic KJ, Bashyal RK, Anthony SG, Chiu V, Shulman B, Rubash HE. Is administratively coded comorbidity and complication data in total joint arthroplasty valid? Clinical orthopaedics and related research 2013; 471(1): 201-5. 11. Brennan SL, Stanford T, Wluka AE, et al. Cross-sectional analysis of association between socioeconomic status and utilization of primary total hip joint replacements 2006-7: Australian Orthopaedic Association National Joint Replacement Registry. BMC musculoskeletal disorders 2012; 13: 63. 12. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987; 40(5): 373-83. 13. Charnley J. Surgery of the hip-joint: present and future developments. Br Med J 1960; 1(5176): 821-6. 14. Cleves MA, Sanchez N, Draheim M. Evaluation of two competing methods for calculating Charlson's comorbidity index when analyzing short-term mortality using administrative data. Journal of clinical epidemiology 1997; 50(8): 903-8.

58

15. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. Journal of clinical epidemiology 1992; 45(6): 613-9. 16. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Medical care 1998; 36(1): 8-27. 17. Fishman PA, Goodman MJ, Hornbrook MC, Meenan RT, Bachman DJ, O'Keeffe Rosetti MC. Risk adjustment using automated ambulatory pharmacy data: the RxRisk model. Medical care 2003; 41(1): 84-99. 18. Francis ML, Scaife SL, Zahnd WE, Cook EF, Schneeweiss S. Joint replacement surgeries among medicare beneficiaries in rural compared with urban areas. Arthritis and rheumatism 2009; 60(12): 3554-62. 19. Gordon M, Stark A, Skoldenberg OG, Karrholm J, Garellick G. The influence of comorbidity scores on re-operations following primary total hip replacement: comparison and validation of three comorbidity measures. The bone & joint journal 2013; 95-B(9): 1184-91. 20. Haynes SR, Lawler PG. An assessment of the consistency of ASA physical status classification allocation. Anaesthesia 1995; 50(3): 195-9. 21. Hollowell J, Grocott MP, Hardy R, Haddad FS, Mythen MG, Raine R. Major elective joint replacement surgery: socioeconomic variations in surgical risk, postoperative morbidity and length of stay. J Eval Clin Pract 2010; 16(3): 529-38. 22. Hossain M, Andrew JG. Is there a difference in perioperative mortality between cemented and uncemented implants in hip fracture surgery? Injury 2012; 43(12): 2161-4. 23. http://www.who.int/classifications/icd/en/ W. 2009. 24. Hunt LP, Ben-Shlomo Y, Clark EM, et al. 90-day mortality after 409,096 total hip replacements for osteoarthritis, from the National Joint Registry for England and Wales: a retrospective analysis. Lancet 2013; 382(9898): 1097-104. 25. Icmje. Uniform requirements for manuscripts submitted to biomedical journals: writing and editing for biomedical publication updated October 2004. Mymensingh medical journal : MMJ 2005; 14(1): 95-119. 26. ICMJE. http://www.icmje.org/recommendations/browse/roles-and-responsibilities/defining-the-role-of-authors-and-contributors.html. 2016. 27. Iezzoni LI. Assessing quality using administrative data. Annals of internal medicine 1997; 127(8 Pt 2): 666-74. 28. Jameson SS, Kyle J, Baker PN, et al. Patient and implant survival following 4323 total hip replacements for acute femoral neck fracture: a retrospective cohort study using National Joint Registry data. The Journal of bone and joint surgery British volume 2012; 94(11): 1557-66. 29. Johnson ML, El-Serag HB, Tran TT, Hartman C, Richardson P, Abraham NS. Adapting the Rx-Risk-V for mortality prediction in outpatient populations. Medical care 2006; 44(8): 793-7. 30. Jones MD, Parry MC, Whitehouse MR, Blom AW. Early death following primary total hip arthroplasty. The Journal of arthroplasty 2014; 29(8): 1625-8.

59

31. Kandala NB, Connock M, Pulikottil-Jacob R, et al. Response to two recent BMJ papers on mortality after hip replacement: comparative modelling study. Bmj 2014; 348: g1506. 32. Kim YH, Kwon OR, Kim JS. Is one-stage bilateral sequential total hip replacement as safe as unilateral total hip replacement? The Journal of bone and joint surgery British volume 2009; 91(3): 316-20. 33. Kork F, Balzer F, Krannich A, Weiss B, Wernecke KD, Spies C. Association of comorbidities with postoperative in-hospital mortality: a retrospective cohort study. Medicine 2015; 94(8): e576. 34. Kragsterman B, Parsson H, Lindback J, Bergqvist D, Bjorck M, Swedish Vascular R. Outcomes of carotid endarterectomy for asymptomatic stenosis in Sweden are improving: Results from a population-based registry. Journal of vascular surgery : official publication, the Society for Vascular Surgery [and] International Society for Cardiovascular Surgery, North American Chapter 2006; 44(1): 79-85. 35. Lalmohamed A, Vestergaard P, de Boer A, Leufkens HG, van Staa TP, de Vries F. Changes in mortality patterns following total hip or knee arthroplasty over the past two decades: a nationwide cohort study. Arthritis Rheumatol 2014; 66(2): 311-8. 36. Learmonth ID, Young C, Rorabeck C. The operation of the century: total hip replacement. Lancet 2007; 370(9597): 1508-19. 37. Lie SA, Engesaeter LB, Havelin LI, Gjessing HK, Vollset SE. Mortality after total hip replacement: 0-10-year follow-up of 39,543 patients in the Norwegian Arthroplasty Register. Acta Orthop Scand 2000; 71(1): 19-27. 38. Ludvigsson JF, Otterblad-Olausson P, Pettersson BU, Ekbom A. The Swedish personal identity number: possibilities and pitfalls in healthcare and medical research. European journal of epidemiology 2009; 24(11): 659-67. 39. Macaulay W, Salvati EA, Sculco TP, Pellicci PM. Single-stage bilateral total hip arthroplasty. The Journal of the American Academy of Orthopaedic Surgeons 2002; 10(3): 217-21. 40. Manning DW, Edelstein AI, Alvi HM. Risk Prediction Tools for Hip and Knee Arthroplasty. The Journal of the American Academy of Orthopaedic Surgeons 2016; 24(1): 19-27. 41. McMinn DJ, Snell KI, Daniel J, Treacy RB, Pynsent PB, Riley RD. Mortality and implant revision rates of hip arthroplasty in patients with osteoarthritis: registry based cohort study. Bmj 2012; 344: e3319. 42. Mundi S, Pindiprolu B, Simunovic N, Bhandari M. Similar mortality rates in hip fracture patients over the past 31 years. Acta orthopaedica 2014; 85(1): 54-9. 43. Ornstein E, Linder L, Ranstam J, Lewold S, Eisler T, Torper M. Femoral impaction bone grafting with the Exeter stem - the Swedish experience: survivorship analysis of 1305 revisions performed between 1989 and 2002. The Journal of bone and joint surgery British volume 2009; 91(4): 441-6. 44. Padyab M, Malmberg G, Norberg M, Blomstedt Y. Life course socioeconomic position and mortality: a population register-based study from Sweden. Scand J Public Health 2013; 41(8): 785-91.

60

45. Parry M, Wylde V, Blom AW. Ninety-day mortality after elective total hip replacement: 1549 patients using aspirin as a thromboprophylactic agent. The Journal of bone and joint surgery British volume 2008; 90(3): 306-7. 46. Parvizi J, Tarity TD, Herz A, et al. Ninety-day mortality after bilateral hip arthroplasty. The Journal of arthroplasty 2006; 21(7): 931-4. 47. Pedersen AB, Baron JA, Overgaard S, Johnsen SP. Short- and long-term mortality following primary total hip replacement for osteoarthritis: a Danish nationwide epidemiological study. The Journal of bone and joint surgery British volume 2011; 93(2): 172-7. 48. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Medical care 2005; 43(11): 1130-9. 49. Quan H, Li B, Couris CM, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol 2011; 173(6): 676-82. 50. Rahman MM, Kopec JA, Sayre EC, et al. Effect of sociodemographic factors on surgical consultations and hip or knee replacements among patients with osteoarthritis in British Columbia, Canada. J Rheumatol 2011; 38(3): 503-9. 51. Ranstam J, Karrholm J, Pulkkinen P, et al. Statistical analysis of arthroplasty data. II. Guidelines. Acta orthopaedica 2011; 82(3): 258-67. 52. Ranstam J, Karrholm J, Pulkkinen P, et al. Statistical analysis of arthroplasty data. I. Introduction and background. Acta orthopaedica 2011; 82(3): 253-7. 53. Ravn H, Bergqvist D, Bjorck M, Swedish Vascular R. Nationwide study of the outcome of popliteal artery aneurysms treated surgically. The British journal of surgery 2007; 94(8): 970-7. 54. Ritter MA, Randolph JC. Bilateral total hip arthroplasty: a simultaneous procedure. Acta Orthop Scand 1976; 47(2): 203-8. 55. Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. Journal of clinical epidemiology 1993; 46(10): 1075-9; discussion 81-90. 56. Saklad M. Grading of patients for surgical procedures. Anesthesiology 1941; 2: 281-4. 57. Shrier I, Platt RW. Reducing bias through directed acyclic graphs. BMC Med Res Methodol 2008; 8: 70. 58. Singh JA, Kundukulam J, Riddle DL, Strand V, Tugwell P. Early postoperative mortality following joint arthroplasty: a systematic review. J Rheumatol 2011; 38(7): 1507-13. 59. Singh JA, Lewallen DG. Ninety-day mortality in patients undergoing elective total hip or total knee arthroplasty. The Journal of arthroplasty 2012; 27(8): 1417-22 e1. 60. Sloan KL, Sales AE, Liu CF, et al. Construction and characteristics of the RxRisk-V: a VA-adapted pharmacy-based case-mix instrument. Medical care 2003; 41(6): 761-74. 61. Soderman P, Malchau H, Herberts P, Johnell O. Are the findings in the Swedish National Total Hip Arthroplasty Register valid? A comparison between

61

the Swedish National Total Hip Arthroplasty Register, the National Discharge Register, and the National Death Register. The Journal of arthroplasty 2000; 15(7): 884-9. 62. Soderman P, Malchau H, Herberts P, Zugner R, Regner H, Garellick G. Outcome after total hip arthroplasty: Part II. Disease-specific follow-up and the Swedish National Total Hip Arthroplasty Register. Acta Orthop Scand 2001; 72(2): 113-9. 63. Stafford GH, Charman SC, Borroff MJ, Newell C, Tucker JK. Total hip replacement for the treatment of acute femoral neck fractures: results from the National Joint Registry of England and Wales at 3-5 years after surgery. Ann R Coll Surg Engl 2012; 94(3): 193-8. 64. Su H, Aharonoff GB, Hiebert R, Zuckerman JD, Koval KJ. In-hospital mortality after femoral neck fracture: do internal fixation and hemiarthroplasty differ? Am J Orthop (Belle Mead NJ) 2003; 32(3): 151-5. 65. Sundararajan V, Henderson T, Perry C, Muggivan A, Quan H, Ghali WA. New ICD-10 version of the Charlson comorbidity index predicted in-hospital mortality. Journal of clinical epidemiology 2004; 57(12): 1288-94. 66. Tarity TD, Herz AL, Parvizi J, Rothman RH. Ninety-day mortality after hip arthroplasty: a comparison between unilateral and simultaneous bilateral procedures. The Journal of arthroplasty 2006; 21(6 Suppl 2): 60-4. 67. Towheed TE, Hochberg MC. Health-related quality of life after total hip replacement. Semin Arthritis Rheum 1996; 26(1): 483-91. 68. Troeng T, Malmstedt J, Bjorck M. External validation of the Swedvasc registry: a first-time individual cross-matching with the unique personal identity number. European journal of vascular and endovascular surgery : the official journal of the European Society for Vascular Surgery 2008; 36(6): 705-12. 69. van Walraven C, Austin PC, Jennings A, Quan H, Forster AJ. A modification of the Elixhauser comorbidity measures into a point system for hospital death using administrative data. Medical care 2009; 47(6): 626-33. 70. Visuri T, Pulkkinen P, Turula KB, Paavolainen P, Koskenvuo M. Life expectancy after hip arthroplasty. Case-control study of 1018 cases of primary arthrosis. Acta Orthop Scand 1994; 65(1): 9-11. 71. Whitehouse SL, Bolland BJ, Howell JR, Crawford RW, Timperley AJ. Mortality following hip arthroplasty--inappropriate use of National Joint Registry (NJR) data. The Journal of arthroplasty 2014; 29(9): 1827-34. 72. Whitmore RG, Stephen JH, Vernick C, et al. ASA grade and Charlson Comorbidity Index of spinal surgery patients: correlation with complications and societal costs. The spine journal : official journal of the North American Spine Society 2014; 14(1): 31-8. 73. World Medical A. World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects. Jama 2013; 310(20): 2191-4. 74. Zoller B, Li X, Sundquist J, Sundquist K. Socioeconomic and occupational risk factors for venous thromboembolism in Sweden: a nationwide epidemiological study. Thromb Res 2012; 129(5): 577-82.

62

63

Epilogue

The answer to my designated mentor’s question as to whether I am planning to die dissertation-less: - No, I’m not.

Acta Universitatis UpsaliensisDigital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Medicine 1309

Editor: The Dean of the Faculty of Medicine

A doctoral dissertation from the Faculty of Medicine, UppsalaUniversity, is usually a summary of a number of papers. A fewcopies of the complete dissertation are kept at major Swedishresearch libraries, while the summary alone is distributedinternationally through the series Digital ComprehensiveSummaries of Uppsala Dissertations from the Faculty ofMedicine. (Prior to January, 2005, the series was publishedunder the title “Comprehensive Summaries of UppsalaDissertations from the Faculty of Medicine”.)

Distribution: publications.uu.seurn:nbn:se:uu:diva-316989

ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2017