345Bulletin of the World Health Organization | May 2005, 83 (5)
Objectives To identify and examine differences in pre-existing morbidity between injured and non-injured population-based cohorts. Methods Administrative health data from Manitoba, Canada, were used to select a population-based cohort of injured people and a sample of non-injured people matched on age, gender, Aboriginal status and geographical location of residence at the date of injury. All individuals aged 18–64 years who had been hospitalized between 1988 and 1991 for injury (International Classification of Diseases, Ninth Edition, Clinical Modification (ICD-9-CM) code 800–995) (n = 21 032), were identified from the Manitoba discharge database. The matched non-injured comparison group comprised individuals randomly selected 1:1 from the Manitoba population registry. Morbidity data for the 12 months prior to the date of the injury were obtained by linking the two cohorts with all hospital discharge records and physician claims. Results Compared to the non-injured group, injured people had higher Charlson Comorbidity Index scores, 1.9 times higher rates of hospital admissions and 1.7 times higher rates of physician claims in the year prior to the injury. Injured people had a rate of admissions to hospital for a mental health disorder 9.3 times higher, and physician claims for a mental health disorder 3.5 times higher, than that of non-injured people. These differences were all statistically significant (P < 0.001). Conclusion Injured people were shown to differ from the general non-injured population in terms of pre-existing morbidity. Existing population estimates of the attributable burden of injury that are obtained by extrapolating from observed outcomes in samples of injured cases may overestimate the magnitude of the problem.
Keywords Wounds and injuries/epidemiology/complications; Comorbidity; Causality; Health services/utilization; Cost of illness; Retrospective studies; Cohort studies; Canada (source: MeSH, NLM). Mots clés Plaies et traumatismes/épidémiologie/complication; Morbidité associée; Causalité; Services santé/utilisation; Coût maladie; Etude rétrospective; Etude cohorte; Canada (source: MeSH, INSERM). Palabras clave Heridas y lesiones/epidemiología/complicaciones; Comorbilidad; Causalidad; Servicios de salud/utilización; Costo de la enfermedad; Estudios retrospectivos; Estudios de cohortes; Canadá (fuente: DeCS, BIREME).
Bulletin of the World Health Organization 2005;83:345-352.
Voir page 351 le résumé en français. En la página 351 figura un resumen en español.
Differences in prevalence of pre-existing morbidity between injured and non-injured populations Cate M. Cameron,1, 2 David M. Purdie,3 Erich V. Kliewer,4 & Roderick J. McClure1
1 School of Population Health, University of Queensland, Mayne Medical School, Herston Road, Herston, Brisbane, Queensland, Australia 4006. Correspondence should be sent to Dr. Cameron at this address (email: cameronc@uq.edu.au). 2 Centre of National Research on Disability and Rehabilitation Medicine, University of Queensland, Queensland, Australia. 3 Northern California Cancer Center, California, USA. 4 CancerCare Manitoba, Manitoba Health, Winnipeg, Canada. Ref. No. 04-011130 (Submitted: 6 January 2004 – Final revised version received: 13 July 2004 – Accepted: 10 November 2004)
Introduction WHO has predicted that injury will be the second leading cause of the world disease burden by the year 2020 (1, 2). A limita- tion in the current process of deriving population estimates of the burden attributable to injury is the failure to take into account pre-existing morbidity. If injured people differ from the general population in terms of pre-existing morbidity, then observed outcomes in injured samples that are attributed to injury may in part be due to pre-existing morbidities rather than to the injury in question.
The Australian Burden of Disease study acknowledged the importance of co-existing conditions in estimating the at- tributable burden of particular conditions (3). Mathers et al. (3) concluded that several methodological issues relating to
comorbidity remain to be addressed if burden of disease models are to be advanced. These issues include how comorbidities affect long-term disability; which comorbidities are relevant; and how to deal with the logistics of modelling large numbers of combinations of comorbidities (3).
Although some attempts have been made to look at differences in health status in patients pre- and post-injury using self-reported retrospective recall (4), investigators using this method acknowledge inevitable biases (5–9). Bias is better managed by ascertaining pre-injury morbidity at a point in time before the injury was sustained. Although rare in injury outcome studies (10, 11), the use of comorbidity indices based on administrative claims data recorded prior to the index event is well established in other fields of research such as cancer, cardiovascular disease and diabetes (12–14).
.352
346 Bulletin of the World Health Organization | May 2005, 83 (5)
Research Pre-existing morbidity in injured and non-injured populations Cate M. Cameron et al.
Global burden of disease estimates currently assume that the distribution of morbidity in the community is independent of injury status. Outcome studies of clinical case series cannot test this assumption and few population-based studies that compare injured and non-injured people have been reported in the literature (10). The present study used administrative health databases to compare the frequency and distribution of morbidity in people in the 12 months prior to their sustain- ing an injury with the health status of the general non-injured population.
Methods Study design The study described in this paper is an examination of the prevalence of pre-existing disease in two samples drawn from administrative health data from Manitoba, Canada. The two samples were originally identified on the basis of exposure to injury, for the purposes of conducting a matched, population- based retrospective cohort study, with a follow-up period of 10 years. The University of Manitoba Research Ethics Board and the Health Information Privacy Committee of Manitoba Health approved this study. Data extractions were completed by Mani- toba Health and all identifying variables were removed from the data before the study investigator was granted access to them.
Data sources The province of Manitoba provides universal health-care cover- age for a population of 1.14 million residents (15). Manitoba Health maintains databases of claims made by health providers for reimbursement of services (hospital, physician and extended- care services), as well as a population registry of those eligible for health coverage (16). Virtually every resident of Manitoba is covered by the provincial health-care plan (15). The databases have been used extensively in health research and are described in detail elsewhere (17, 18).
Setting and participants A cohort of injured people (n = 21 032) was identified as all persons aged 18–64 years resident in the province of Manitoba, who had been hospitalized for treatment of an injury between 1 January 1988 and 31 December 1991. The cohort members included all individuals who had an International Classification of Diseases, Ninth edition, Clinical Modification (ICD-9-CM) code 800–995 (excluding late effects from injury 905–909, and allergies from within 995), in the first or second diagnostic fields. For individuals who had more than one injury-related hospital admission during the study period, the first admission was designated as the index case record.
For each injured subject, a non-injured person was ran- domly selected from the Manitoba population registry and matched on age, gender, aboriginal status and geographical location of residence at the date of admission of the index case. Excluded from both the injured and non-injured cohorts were residents of nursing homes, patients in extended hospital care and people who had not been resident in the province for 12 months prior to the admission date of the index record (for the purposes of extracting baseline data).
Health service utilization and measurement of comorbidity Information on health service utilization was extracted from hospital discharge data and physician services claims for both
the injured and non-injured cohorts. Dates of admissions, services, treatment, diagnostic and discharge information were extracted for the 12-month period prior to the date of the index record for all matched pairs (19). Physician claims were re- stricted to ambulatory claims, including outpatient, emergency department (where available) and visits to medical practitioners in the community (20, 21). For individuals with multiple hos- pital records associated with a single episode (e.g. when inter- hospital transfers or readmissions had taken place), they were integrated into one summary record. Frequency and types of pre-existing comorbid conditions for the two cohorts were determined from the extracted hospital and physician claims during the pre-injury period. For the purpose of identifying pre-existing comorbidities, only the primary diagnostic field was used, and conditions were categorized according to the 18 disease chapters of ICD-9-CM. The number of comorbid conditions was counted by summing the number of different disease categories present. Data on comorbid conditions from physician claims were derived using the same method.
The Dartmouth-Manitoba version of the Charlson Comorbidity Index (CCI) (22) was also used to quantify pre- existing comorbidity. The CCI was computed for the injured cohort based on 12 months of hospital data for the year preced- ing the index injury record. For the non-injured cohort, the CCI was calculated for the 12-month period prior to date of injury of the matched case. If no comorbid conditions were found, or where no hospital records existed, the CCI score was set to zero.
Injury classification The injured were analysed as the total cohort and by the nature of injury codes (ICD-9-CM 800–995). Seven subgroups (brain injury, spinal injury, burns, long-bone fractures, poisonings, internal injuries and other) were created across subchapter headings to enable more comprehensive examination of those injury types commonly studied in injury outcomes research. ICDMAP-90 © software from Johns Hopkins University was used to generate an Injury Severity Score (ISS) for injured cases. These ISS scores were grouped into minor (ISS 1–8), moderate (ISS 9–15) and severe (ISS 16) categories in accordance with the convention (23). ICDMAP-90 © maps a severity score for only a proportion of the total Injury and Poisonings ICD-9-CM codes. Therefore not all injured cases were scored.
Analysis Analysis of data involved univariate and bivariate statistics. The statistical significance of differences between groups and subgroups was assessed by chi-squared statistics for categorical data and with the Mann–Whitney U test for continuous data because of non-normal distributions. Rate ratio confidence intervals were adjusted for matching variables using Poisson regression methods. All tests were two sided with a 5% level of significance. Rates of health service utilization were reported using person-years (PYs) of exposure time. Analysis was con- ducted using SAS version 8.2 statistical software.
Results Sample characteristics The characteristics of the 21 032 injured cases are shown in Table 1. The mean age at the start of follow-up for both injured and non-injured cohorts was 35.7 years. Males represented
347Bulletin of the World Health Organization | May 2005, 83 (5)
Research Cate M. Cameron et al. Pre-existing morbidity in injured and non-injured populations
almost two-thirds of the cohort and more injuries occurred in individuals in the younger age categories (18–34 years). Of the injury subgroups, fractures of long bones (12%), poisonings (10.3%) and brain injury (6.1%) were the most common. An ISS was generated for 62% of the total injured (n = 17 002). Over 85% of scored cases were of minor severity, 10% of mod- erate severity and almost 4% were classified as major injuries.
Pre-existing morbidity for injured sample Members of the injured cohort had a 1.9 times higher rate of hospital admissions in the 12 months before the date of the index record than members of the non-injured cohort (injured 276/1000 PYs and non-injured 148/1000 PYs). The mean num- ber of hospital admissions per person in the pre-injury year was 0.12 in the injured (range 0–19) and 0.09 in the non-injured cohort (range 0–12). The mean length of stay in hospital dif- fered significantly between the two cohorts in the pre-injury year (Table 2). Injured people were less likely to be admitted as day patients (28.6% versus 36.3%) and more likely to have a total length of stay greater than 14 days (13.3% versus 5.7%).
The injured cohort had a 1.7 times higher rate of phy- sician claims in the pre-injury period (670/100 PYs, average 3.9 claims per person) than the non-injured cohort (387/100 PYs, average 2.2 claims per person). A greater percentage of the injured cohort (65%) than of the non-injured cohort (48%) had three or more physician claims in the 12-month pre-injury period (Table 2).
Significantly more people from the injured cohort (5.9%) than from the non-injured cohort (1.2%) had CCI scores of 1 or more (Table 3). Overall, a greater percentage of injured people had comorbidities, and they had more comorbidities per person than the members of the non-injured cohort, based on hospitalizations and physician claims. Injured people had an average of 2.2 different conditions (range 0–14) for which they had consulted a physician in the pre-injury year, whereas in the non-injured people this figure was 1.5 (range 0–13). There were 1498 people in the injured group and 423 in the non-injured cohort, who had a moderate or severe pre-existing mental health condition, as indicated by health service use in the pre-injury period.
The members of the injured cohort had higher rates of hospital admissions and physician claims for all causes in the pre-injury period than those in the non-injured cohort (Table 4 and Table 5). The rate ratios were 1.5 or greater for 14 of the 18 chapters for hospital admissions and for six chapters for physician claims. When ranked by rate ratio, for both hospital admissions and physician claims, the greatest differences were seen for mental health disorders and previous injuries. Injured people had a 9.3 times higher rate of admissions to hospital for a mental health disorder (injured 42/1000 PYs and non-injured 4.5/1000 PYs) and a 3.5 times higher rate of mental health phy- sician claims (injured 98/100 PYs) than the non-injured people (28/100 PYs). Mental disorders were the most frequent cause of all hospital admissions for the total injured cohort. Almost half of the mental health admissions were for alcoholic psychoses, affective psychoses and schizophrenic disorders. Over 80% of all mental health physician claims for the injured cohort were for “non-psychotic or personality disorders”, more specifically for such conditions as panic, anxiety or depression.
During the pre-injury period, the injured cohort had a 3.7 times higher rate of admissions to hospital for a previous injury (injured 9.9/1000 PYs and non-injured 2.7/1000 PYs)
Table 1. Demographic characteristics for injured and non- injured cohorts, injury subgroups and injury severity scores at the time of the case index injury admission
Injured Non-injured (n = 21 032) (n = 21 032)
n % n %
Gender Male 13 441 63.9 13 441 63.9 Female 7 591 36.1 7 591 36.1
Age in yearsa 18–24 5 410 25.7 5 422 25.7 25–34 6 014 28.6 5 990 28.6 35–44 3 959 18.8 3 972 18.8 45–54 2 805 13.3 2 799 13.3 55–64 2 844 13.5 2 849 13.5
Place of residencea Urban 8 687 41.3 8 799 41.9 Rural 8 167 38.8 8 208 39.0 Remote 4 178 19.9 4 025 19.1
Injury subgroups Brain injury 1 290 6.1 Spinal injury 95 0.5 Burns 524 2.5 Fractures of long bones 2 515 12.0 Poisonings 2 169 10.3 Internal injuries 593 2.8 Other injuries 13 846 65.8
Injury Severity Score (ISS)b Minor (ISS 1–8) 14 599 85.9 Moderate (ISS 9–15) 1 746 10.3 Severe (ISS 16) 657 3.8
a Age-matched on year of birth and place of residence on partial postcode, thus there are small differences in actual numbers of injured and non-injured. b ISS not computed for 4030 cases.
and a 2.7 times higher rate of physician claims for injury-related matters (117/100 PYs) than the non-injured group (43/100 PYs). Nineteen per cent of the previous injury-related admis- sions for the injured cohort (and 10% for the non-injured) were for self-inflicted harm or suicide attempts by poisoning. Nineteen per cent of the previous physician injury claims for the injured group (and 28% for the non-injured) were for sprains and strains.
Pre-existing morbidity by injury type Health service use in the 12-month pre-injury period was similar across the injury types analysed. However, members of the injured cohort hospitalized for poisonings were notable in that they only accounted for 10% of the injured, but accounted for 24% of all admissions and 20% of all physician claims in the pre-injury period. Furthermore, this group had a 59 times higher rate of hospital admissions (95% confidence interval (CI), 27.75–123.64) and an 11 times higher rate of physician claims (95% CI, 10.38–12.0) for mental health disorders than their matched counterparts in the uninjured group. Over 50% of the hospital admissions for mental health problems and 78% of physician claims for the poisonings group in the pre-injury
348 Bulletin of the World Health Organization | May 2005, 83 (5)
Research Pre-existing morbidity in injured and non-injured populations Cate M. Cameron et al.
Table 2. Hospital admissions, length of stay and physician claims in the 12-month period prior to the index admission for injured and non-injured cohorts
Injured (21 032) Non-injured (21 032) Significance
n % n %
Total hospital admissions 5 804 100 3 111 100 Admissions per person No admissions 17 316 82.33 18 665 88.75 1 admission 2 569 12.21 1 879 8.93 P < 0.0001a
2 admissions 698 3.32 360 1.71 3 admissions 236 1.12 78 0.37 4 or more admissions 213 1.01 50 0.24 Mean number of admissions (range)c 0.12 (0–19) 0.09 (0–12) P < 0.0001b
Total length of stay in hospital (days) Day patient 1 062 28.58 859 36.29 1–2 625 16.82 363 15.34 3–4 575 15.47 473 19.98 P < 0.0001a
5–7 543 14.61 329 13.90 8–14 416 11.19 208 8.79 15–29 281 7.56 79 3.34 30 days or more 214 5.76 56 2.37 Mean length of stay in days (range)c 0.36 (0–248) 0.18 (0–258) P < 0.0001b
Total physician claims 140 986 100 81 302 100 Claims per person No claims 2 746 13.06 5 137 24.42 1–2 claims 4 685 22.28 5 864 27.88 3–5 claims 5 328 25.33 5 082 24.16 P < 0.0001a
6–10 claims 4 342 20.64 3 246 15.43 11–20 claims 2 736 13.01 1 396 6.64 21–50 claims 1 064 5.06 286 1.36 51 or more claims 131 0.62 21 0.10 Mean number of claims (range)c 3.9 (0–295) 2.2 (0–127) P < 0.0001b
a Determined by χ2 statistics. b Determined by Mann–Whitney U test. c Geometric mean calculated due to non-normal distributions.
Health service use for the 12 months prior to injury
year were for non-psychotic, neurotic or personality disorders — often for specific anxiety or depressive disorders. The poi- sonings group had significantly more hospital admissions in the pre-injury year for previous injuries and poisonings (rate ratio (RR) = 11.5; 95% CI, 4.14–31.95), circulatory diseases (RR = 6.2; 95% CI, 3.22–11.52) and ill-defined conditions (RR = 4.7; 95% CI, 2.82–7.26) than the matched subgroup. This injury group also had higher rates of physician claims for ill-defined conditions (RR = 3.4; 95% CI, 3.12–3.60), neoplasms (RR = 2.9; 95% CI, 2.27–3.77), digestive diseases (RR = 2.7; 95% CI, 2.39–2.96), than the matched subgroup. Excluding poisonings from the analysis had little effect on the statistical significance of differences between health service utilization by the members of the injured and non-injured cohorts during the pre-injury period.
Discussion To the best of our knowledge, this is the first study to quan- tify a difference in pre-existing morbidity between injured and non-injured populations. Compared to non-injured people, injured people were almost five times as likely to have a CCI score of one or more, were admitted to hospital at almost twice as often, and had 1.7 times the rate of physician claims. The
injured cohort had fewer members with no comorbid condi- tions and significantly more members with greater numbers of different conditions present than did the matched sample from the general population. The results of this study support the concern that current burden of injury estimates may be inaccurate if the influence of pre-existing ill-health is not taken into account.
Although few studies reported in the literature have been similar in scope and methods to the current study (3), one key population-based study of elderly people with and without hip fractures, did find similar results (10). This study found that, prior to their injury, the study subjects had significantly higher comorbidities using the CCI, higher disability measures and were more likely to be residents of nursing homes than controls (10).
Several studies of clinical injury outcome have measured comorbidities using a number of different methods, mostly rely- ing on the diagnostic fields of the index admission record (7, 8, 24). These studies have found a consistent association between the presence of comorbid conditions and increased risk of mor- tality following injury. Smaller and less consistent associations were found in the few studies that have considered comorbidity and non-fatal outcomes (25–27).
349Bulletin of the World Health Organization | May 2005, 83 (5)
Research Cate M. Cameron et al. Pre-existing morbidity in injured and non-injured populations
Table 3. Comorbid conditions in the 12-month period prior to the index admission for injured and non-injured cohorts
Injured (21 032) Non-injured (21 032) Significance
n % n %
Charlson Comorbidity Index No comorbidity 19 797 94.12 20 778 98.79 P < 0.0001a
1 or more comorbidity 1 235 5.87 254 1.21
Comorbid conditions per person (based on hospital admissions) No comorbidities 17 316 82.33 18 665 88.75 1 comorbidity 3 004 14.28 2 110 10.03 P < 0.0001a
2 comorbidities 571 2.71 224 1.06 3 or more comorbidities 141 0.67 33 0.16 Mean number of conditions (range)c 0.15 (0–6) 0.09 (0–6) P < 0.0001b
Comorbid conditions per person (based on physician claims) No comorbidities 2 746 13.06 5 137 24.42 1 comorbidity 3 838 18.25 4 750 22.58 2 comorbidities 4 018 19.10 3 959 18.82 P < 0.0001a
3 comorbidities 3 341 15.89 2 904 13.81 4 comorbidities 2 604 12.38 1 944 9.24 5 or more comorbidities 4 485 21.32 2 338 11.12 Mean number of conditions (range)c 2.2 (0–14) 1.5 (0–13) P < 0.0001b
a Determined by χ2 statistics. b Determined by Mann-Whitney U test. c Geometric mean calculated due to non-normal distributions.
Comorbid conditions for the 12 months prior to injury
Table 4. Hospital admissions per 1000 person-years for all ICD-9-CM disease chapters during the 12-month period prior to the injury, for the injured and matched non-injured cohort, ranked by rate ratios
Rate of hospital admissions per 1000 PYsa
ICD-9-CM chapters Injured Non-injured Adjusted rate ratiob 95% confidence n = 21 032 n = 21 032 intervals
Mental health disorders 42.03 4.52 9.31 7.53–11.50 Injury and poisonings 9.98 2.71 3.68 2.73–4.90 Blood diseases 1.76 0.52 3.36 1.71–6.57 Endocrine and metabolic 7.56 2.71 2.79 2.05–3.75 Musculoskeletal diseases 18.59 7.13 2.61 2.16–3.15 Nervous system diseases 12.31 5.23 2.35 1.88–2.94 Ill-defined conditions 18.50 8.04 2.30 1.91–2.74 Respiratory diseases 16.78 8.46 1.98 1.65–2.36 Circulatory diseases 13.12 7.70 1.70 1.40–2.07 Supplementary classification 15.12 9.56 1.58 1.32–1.89 Congenital anomalies 0.81 0.52 1.55 0.73–3.31 Skin diseases 12.41 8.04 1.54 1.27–1.87 Digestive diseases 32.05 21.11 1.52 1.34–1.71 Infectious diseases 3.57 2.38 1.50 1.04–2.13 Genitourinary diseases 22.97 15.74 1.46 1.26–1.67 Complications of pregnancy 36.80 33.00 1.12 1.01–1.24c
Neoplasms 11.60 10.56 1.10 0.91–1.32 Conditions of perinatal period 0.00 0.00 NA NA a PYs = person years. b Adjusted for age, gender and place of residence. c Unadjusted confidence interval.
350 Bulletin of the World Health Organization | May 2005, 83 (5)
Research Pre-existing morbidity in injured and non-injured populations Cate M. Cameron et al.
Table 5. Physician claims per 100 person-years for all ICD-9-CM disease chapters during the 12-month period prior to the injury, for the injured and matched non-injured cohort, ranked by rate ratios
Rate of physician claims per 100 PYsa
ICD-9-CM chapters Injured Non-injured Adjusted rate ratiob 95% confidence n = 21 032 n = 21 032 intervals
Mental health disorders 98.15 28.05 3.50 3.40–3.61 Injury and poisonings 117.41 43.23 2.72 2.65–2.78 Ill-defined conditions 61.21 33.72 1.82 1.76–1.87 Musculoskeletal diseases 63.74 36.18 1.76 1.71–1.81 Digestive diseases 30.29 18.72 1.62 1.55–1.68 Blood diseases 3.48 2.27 1.53 1.36–1.72 Nervous system diseases 43.69 30.85 1.42 1.37–1.46 Congenital anomalies 0.67 0.47 1.41 1.09–1.83 Respiratory diseases 71.26 51.52 1.38 1.35–1.42 Endocrine and metabolic 23.67 17.14 1.38 1.32–1.44 Genitourinary diseases 40.66 29.67 1.37 1.33–1.42 Skin diseases 33.17 24.85 1.33 1.29–1.38 Infectious diseases 15.37 11.72 1.31 1.24–1.38 Neoplasms 5.75 4.52 1.27 1.17–1.39 Circulatory diseases 29.11 24.10 1.21 1.16–1.25 Supplementary classification 31.49 28.44 1.11 1.07–1.15 Complications of pregnancy 1.20 1.08 1.11 0.92–1.32c
Conditions of perinatal period 0.03 0.04 0.75 0.26–2.18
a PYs = person-years. b Adjusted for age, gender and place of residence. c Unadjusted confidence interval.
This study identified two areas of importance for estimat- ing the attributable burden of injury. The first is that injured people have significantly more hospital admissions and physi- cian claims prior to their injury, across almost every disease classification, than do the general population. The differential was most pronounced for hospitalizations for which 14 of the 18 chapters had rate ratios of 1.5 or greater, whereas six of the 18 chapters for physician claims had rate ratios of 1.5 or greater. When ranked by the rate ratios, two disease chapters, mental health disorders and previous injuries, consistently had the highest rates of hospital admission and physician claims. The finding regarding mental health is consistent with a previous study of trauma patients, which found that more than 30% of trauma patients were affected by psychiatric morbidity before the trauma event, and that psychiatric conditions represented 18% of the total morbidity (9). Outcome studies of the mental health consequences of injury should consider controlling for pre-existing conditions before ascribing mental health disorders as a consequence of the injury.
Second, there was no single injury subgroup that ac- counted for the overall differences in the prevalence of pre- existing disease. However, the poisonings subgroup had greater absolute and relative morbidity in almost all ICD-9-CM dis- ease chapters and had the highest rate ratio for mental health admissions when compared with the findings in non-injured counterparts. The poisonings subgroup tended to be younger, female and had made greater use of health services for treatment of non-psychotic, neurotic or personality disorders. Nevertheless, excluding this group from the analysis had little effect on the overall findings. Health service use following an admission for poisoning, if considered solely as a consequence of that injury
admission does not take account of the high baseline service use that would most likely have continued irrespective of the poisoning event.
Aspects of the methodology need to be considered as pos- sible study limitations. First, “health service use” was used as a proxy for health status. Although previous studies conducted in this study population have supported the validity of this practice (21, 28), it does entail several assumptions about the equity of access to health services, patterns of referral, and service delivery being dependent on health status and consistent across all sub- sections of the community. Second, the generalizability of the findings may be compromised by injured cases being defined as those hospitalized during the study period. This excludes injuries that did not lead to hospitalization and pre-hospital fa- talities, and may be biased by issues of service access and referral pathways (29). Third, is the question of whether incident cases could be accurately identified from the administrative databases used to select the injured cohort. Of the 21 032 people in the injured sample only 60 had been admitted in the previous 12 months for an injury with the same ICD-9 injury group code. Of these, 24 were cases of poisoning with repeat admissions mostly occurring after a number of months, and are likely to have been new episodes rather than repeat admissions for the original injury, leaving only 36 cases (i.e. 0.17%) which could potentially have been misclassified as incident cases. Finally, examination of the six categories of injury type, while justified on the basis that they are the groups most relevant to injury out- comes research, accounts for only 35% of all the cases. Further breakdown of the “other” category for the purposes of analysis created a number of small categories and did not contribute information relevant to the aims of the study.
351Bulletin of the World Health Organization | May 2005, 83 (5)
Research Cate M. Cameron et al. Pre-existing morbidity in injured and non-injured populations
There are many strengths which set this study apart from previous studies that have included consideration of comorbidity in their examination of injury outcomes. The injured group in this study is a population-based incidence sample and as such, avoids the biases inherent in most clinical study samples. This is one of the few studies to have included a population-based non-injured com- parison group. The sample size of the injured group was sufficiently large to enable a considerable degree of analysis of injury subgroups, which will provide a rich context for future work in the field. Of principal value has been the ability to measure health status prior to injury in the injured sample, and previous health status in the non- injured group. Through the use of administrative data, analysis of the 12-month pre-injury period circumvents the recall bias noted previously, which has hampered attempts to identify pre-injury
health status. This study has shown that clarifying the contribu- tion of pre-existing morbidity, to enable the refined calculation of population attributable-burden estimates, is vitally important in burden-of-disease approaches to public health policy. O
Acknowledgements We gratefully acknowledge the provision of the data used in this study by Manitoba Health, Winnipeg, Canada. We also wish to thank Mr André Wajda for his invaluable assistance with data extraction and project consultation. This research was funded by the Centre of National Research on Disability and Rehabili- tation Medicine, University of Queensland, Australia.
Competing interests: none declared.
Résumé
Comparaison de la morbidité pré-existante entre des populations de personnes ayant été victimes d’un traumatisme et ne l’ayant pas été Objectif Identifier et étudier les différences de morbidité pré- existante entre des cohortes en population de personnes ayant été victimes d’un traumatisme et ne l’ayant pas été. Méthodes Les auteurs ont utilisé les données sanitaires administratives du Manitoba, Canada, pour sélectionner une cohorte en population de personnes ayant subi un traumatisme et un échantillon de personnes n’en ayant pas subi, appariées selon l’âge, le sexe, l’origine autochtone ou non et le lieu de résidence à la date du traumatisme. Ils ont identifié, à partir de la base de données de décharge hospitalière du Manitoba, tous les individus âgés de 18 à 64 ans ayant fait l’objet d’une hospitalisation pour traumatisme entre 1988 et 1991 [Classification internationale des maladies, neuvième édition, Modification clinique (ICD-9-CM) codes 800 à 995] (n = 21 032). Les auteurs avaient constitué le groupe de comparaison, composé de personnes n’ayant pas subi de traumatisme appariées, en choisissant au hasard, dans un rapport 1/1, des individus dans le registre de population du Manitoba. Ils ont obtenu les données de morbidité pour les 12 mois précédant la date du traumatisme en mettant en relation les deux cohortes avec l’ensemble des registres de décharge hospitalière et des demandes de
remboursement de soins médicaux. Résultats Par rapport au groupe de personnes n’ayant pas subi de traumatisme, les individus victimes d’un traumatisme présentaient des valeurs plus fortes de l’indice de comorbidité de Charlson, des taux d’hospitalisation 1,9 fois plus élevés et des taux de remboursement de soins médicaux 1,7 fois plus élevés dans l’année précédant le traumatisme. Les personnes ayant subi un traumatisme présentaient un taux d’hospitalisation pour troubles mentaux 9,3 fois plus élevé et des taux de demande de remboursement concernant des troubles mentaux 3,5 fois plus élevés que les personnes n’ayant pas été victimes d’un traumatisme. Ces différences étaient toutes statistiquement significatives (p < 0,001). Conclusion Les auteurs ont montré que les personnes victimes de traumatisme différaient de la population générale n’ayant pas subi de traumatisme par la morbidité pré-existante. Les estimations en population disponibles de la charge de traumatisme attribuable à la morbidité pré-existante, obtenues par extrapolation des résultats observés sur des échantillons de cas de traumatismes, peuvent surestimer l’ampleur du problème.
Resumen
Diferencias en la prevalencia de morbilidad previa entre poblaciones de traumatizados y no traumatizados Objetivo Identificar y analizar las diferencias en la morbilidad preexistente entre cohortes poblacionales de personas traumatizadas y no traumatizadas. Métodos Se usaron los datos administrativos sanitarios de Manitoba (Canadá) para seleccionar una cohorte poblacional de personas traumatizadas y una muestra de personas no traumatizadas apareadas por edad, sexo, aboriginalidad y lugar geográfico de residencia en la fecha del traumatismo. Se identificó en la base de datos de altas de Manitoba a todos los individuos de 18 a 64 años que habían sido hospitalizados entre 1988 y 1991 a causa de traumatismos (Clasificación Internacional de Enfermedades, 9ª revisión, Modificación Clínica (CIE-9-MC) códigos 800–995) (n = 21 032). El grupo de comparación de no traumatizados apareados estaba integrado por individuos seleccionados al azar, en proporción 1:1, a partir del registro de población de Manitoba. Se obtuvieron los datos de morbilidad correspondientes a los 12 meses previos a la fecha del traumatismo, para lo cual se relacionaron las dos cohortes
con todos los registros de altas hospitalarias y las facturas de los médicos. Resultados En comparación con el grupo de no traumatizados, para las personas traumatizadas se hallaron puntuaciones mayores del Índice de Comorbilidad de Charlson, tasas de ingreso hospitalario 1,9 veces mayores, y tasas de facturación de médicos 1,7 veces superiores durante el año previo al traumatismo. Las personas traumatizadas presentaban tasas de ingreso hospitalario por trastornos de salud mental 9,3 veces superiores, y en su caso las facturas médicas por tales trastornos eran 3,5 veces mayores que las de los no traumatizados. Todas esas diferencias eran estadísticamente significativas (P < 0,001). Conclusión Las personas traumatizadas diferían de la población no traumatizada en cuanto a la morbilidad preexistente. Las actuales estimaciones poblacionales de la carga atribuible de traumatismos, obtenidas extrapolando a partir de los resultados observados en muestras de personas traumatizadas, podrían estar sobrestimando la magnitud del problema.
352 Bulletin of the World Health Organization | May 2005, 83 (5)
Research Pre-existing morbidity in injured and non-injured populations Cate M. Cameron et al.
References 1. Murray CJ, Lopez AD. The global burden of disease: a comprehensive assessment of mortality and morbidity from diseases, injuries and risk factors in 1990 and projected to 2020. Cambridge (MA): Harvard School of Public Health; 1996. 2. Stone DH, Jarvis S, Pless B. The continuing global challenge of injury. BMJ 2001;322:1557-8. 3. Mathers CD, Vos ET, Stevenson CE, Begg SJ. The burden of disease and injury in Australia. Bulletin of the World Health Organization 2001;79:1076-84. 4. Ameratunga SN, Norton RN, Whitlock G, Macmahon S, Coggan, C, Jackson RT, et al. The New Zealand Blood Donors’ Health Study: baseline findings of a large prospective cohort study of injury. Injury Prevention 2002;8:66-9. 5. Whitlock JA, Hamilton BB. Functional outcome after rehabilitation for severe traumatic brain injury. Archives of Physical Medicine and Rehabilitation 1995; 76:1103-12. 6. Morris JA Jr, MacKenzie EJ, Edelstein SL. The effect of preexisting conditions on mortality in trauma patients. JAMA 1990; 263:1942-6. 7. Sacco WJ, Copes WS, Bain LW Jr, MacKenzie EJ, Frey CF, Hoyt DB, et al. Effect of preinjury illness on trauma patient survival outcome. Journal of Trauma 1993; 35:538-42; discussion 542-3. 8. Milzman DP, Boulanger BR, Rodriguez A, Soderstrom CA, Mitchell KA, Magnant CM. Pre-existing disease in trauma patients: a predictor of fate independent of age and injury severity score. Journal of Trauma 1992;32:236-43; discussion 243-4. 9. Wardle TD. Co-morbid factors in trauma patients. British Medical Bulletin 1999;55:744-56. 10. Leibson CL, Tosteson AN, Gabriel SE, Ransom JE, Melton LJ. Mortality, disability, and nursing home use for persons with and without hip fracture: a population-based study. Journal of the American Geriatrics Society 2002;50:1644-50. 11. Rozental TD, Branas CC, Bozentka DJ, Beredjiklian PK. Survival among elderly patients after fractures of the distal radius. Journal of Hand Surgery 2002;27:948-52. 12. Klabunde CN, Warren JL, Legler JM. Assessing comorbidity using claims data: an overview. Medical Care 2002;40 Suppl:IV 26-35. 13. Reid BC, Alberg AJ, Klassen AC, Rozier RG, Garcia I, Winn DM, et al. A comparison of three comorbidity indexes in a head and neck cancer population. Oral Oncology 2002;38:187-94. 14. Schneeweiss S, Maclure M. Use of comorbidity scores for control of confounding in studies using administrative databases. International Journal of Epidemiology 2000;29:891-8. 15. Bernstein CN, Blanchard JF, Leslie W, Wajda A, Yu BN. The incidence of fracture among patients with inflammatory bowel disease. A population- based cohort study. Annals of Internal Medicine 2000;133:795-9.
16. Young TK, Kliewer E, Blanchard J, Mayer T. Monitoring disease burden and preventive behavior with data linkage: cervical cancer among Aboriginal people in Manitoba, Canada. American Journal of Public Health 2000;90:1466-8. 17. Roos LL, Nicol JP. A research registry: uses, development, and accuracy. Journal of Clinical Epidemiology 1999;52:39-47. 18. Burchill C, Roos LL, Fergusson P, Jebamani L, Turner K, Dueck S. Organizing the present, looking to the future: an online knowledge repository to facilitate collaboration. Journal of Medical Internet Research 2000;2:E10. 19. Finn J, Holman D, Jacobs I. The use of linked ambulance data to estimate the effect of comorbidity on determinants and outcomes of out-of-hospital cardiac arrest in Perth, Western Australia. Symposium on Health Data Linkage: Its value for Australian health policy development and policy relevant research, Sydney, Australia, 20–21 March, 2002. Adelaide, South Australia: Adelaide University; 2002. p. 72-77. Available from: http:// www.publichealth.gov.au/symposium.htm 20. Tataryn DJ, Roos NP, Black CD. Utilization of physician resources for ambulatory care. Medical Care 1995;33 Suppl:DS84-99. 21. Cohen MM, MacWilliam L. Measuring the health of the population. Medical Care 1995;33 Suppl:DS21-42. 22. Roos LL, Stranc L, James RC, Li J. Complications, comorbidities, and mortality: improving classification and prediction. Health Services Research 1997;32:229-38; discussion 239-42. 23. Stevenson M, Segui-Gomez M, Lescohier I, Di Scala C, McDonald-Smith G. An overview of the Injury Severity Score and the New Injury Severity Score. Injury Prevention 2001;7:10-3. 24. Morris JA Jr, MacKenzie EJ, Damiano AM, Bass SM. Mortality in trauma patients: the interaction between host factors and severity. Journal of Trauma 1990;30:1476-82. 25. Rochon PA, Katz JN, Morrow LA, McGlinchey-Berroth R, Ahlquist MM, Sarkarati M, et al. Comorbid illness is associated with survival and length of hospital stay in patients with chronic disability. A prospective comparison of three comorbidity indices. Medical Care 1996;34:1093-101. 26. Richmond TS. An explanatory model of variables influencing postinjury disability. Nursing Research 1997;46:262-9. 27. Nordin M, Hiebert R, Pietrek M, Alexander M, Crane M, Lewis S. Association of comorbidity and outcome in episodes of nonspecific low back pain in occupational populations. Journal of Occupational and Environmental Medicine 2002;44:677-84. 28. Roos NP, Black CD, Frohlich N, Decoster C, Cohen MM, Tataryn DJ, et al. A population-based health information system. Medical Care 1995;33 Suppl:DS13-20. 29. Langley J, Stephenson S, Cryer C, Borman B. Traps for the unwary in estimating person based injury incidence using hospital discharge data. Injury Prevention 2002;8:332-7.