Prevalence of claims-based recurrent low back pain in a Canadian population: A secondary analysis of an administrative database

A secondary analysis of medical administrative data, obtained from the databases of the Régie de l’ Assurance Maladie du Québec (RAMQ), was used to perform a retrospective population-based cohort study. In the province of Quebec, the RAMQ is the government agency responsible for administering the provincial health plan, which came into effect in 1970. This plan covers medical services for Quebec residents. The RAMQ centralizes billing and service information from physicians and hospitals. In this study, the physician claims database for Quebec was used, providing information on the patient’s identification, the date of service and the primary diagnosis of the visit (four-digit International Classification of Diseases, 9^th Revision, or ICD-9, codes). Each Quebec resident is assigned a unique health insurance number for identification purposes, and it is further encrypted by the RAMQ for confidentiality reasons. Demographic estimates by sex and age for the province of Quebec were obtained from Statistics Canada[25]. The local Institutional Ethics Board and the Commission d’accès à l’information du Québec approved this study.

The cohort was selected among Quebec residents, aged 18 and older, who consulted a physician (primary care or specialist) in an ambulatory care facility (emergency, urgent care clinic or outpatient clinic) for a low back pain condition between January 1, 1999 and December 31, 2008. In this study, we considered the older population as being ageg 65 or older. 401 264 adult patients were selected using the physician claims database for “claims-based recurrent low back pain” if they had at least three identical LBP diagnoses (3-digit ICD-9 codes 721, 722, 724 or 739) within a period of 365 days, with at least one of the diagnoses in the year under study.

The annual prevalence reports the proportion of patients identified with recurrent claims-based LBP for the estimated adult population in the province of Quebec for every year between 2000 and 2007. The annual frequency of ICD-9 codes was also reported according to a classification proposed by Cherkin and colleagues, which is based on clinical categories of mechanical low back problems[23]. Extraction of the data was performed with SAS (version 9.2; SAS Institute Inc, NC, USA). Graphing, linear regressions and relative risks were prepared with GraphPad (version 5.d; Graphpad Software Inc., CA, USA). Confidence intervals (99%) were determined with the Wilson score interval method.

Our data showed that adult men in the age categories corresponding to the labor force represented a peak of recurrent medical visits for LBP. This is in accordance with a previous report that showed a predominance of LBP episodes in the 45-54 age category, representing 24.8% of a 15 567 patient cohort[21]. However, our results also showed that women behave differently than men with respect to the number of recurrent visits for a low back pain condition. The female cohort showed a sharper bi-phasic distribution, with a large increase in the peak of visits after the retirement age in 2007 compared to 2000. In line with this, a 2002 US national health survey on a sample of 31 000 respondents reported an increase from 15%, in the 18-24 age category, to 19.7% in the aging population (>65 years)[37]. Overall, women were also significantly more prevalent than men in that study, although the authors did not specifically state whether they were older women. Furthermore, our data also pointed out that men were at greater risk than women in age categories corresponding to the labor force. This observation was noted for early 2000, and it leveled out towards 2007. This could be the result of many factors including a transition away from professions involving manual labor in Quebec in this 8-year period, or an improvement in the prevention and security rules at work. Our results also revealed a higher prevalence for claims-based recurrent LBP in 2000 in men compared to women. However, in 2007, the prevalence of LBP in both sexes became equal at 1.33%. While many studies report a higher prevalence for chronic pain in women[38, 39], a recent web-based survey study in the US revealed no difference in the prevalence between men and women for chronic LBP[7]. Also, we noted that for both sexes, there was a drop in the prevalence in the 60-64 age category in comparison to the peak in the early fifties. In Canada, the mean retirement age is 61 years of age[40], which is in line with the prevalence decrease in the 60-64 age category. It has been reported that the frequency of severe back pain increases with age, but that adults in age categories under 65 years of age are most affected by benign and mixed back pain[41]. In addition, based on a decennial survey in France, active men and women of pre-retirement age showed a higher prevalence of LBP of more than 30 days, which progressively decreased past the retirement age and stabilized around the age category of 70-74[42]. However, our data revealed that the prevalence of claims-based recurrent LBP increased in the elderly, with older women being more at risk than older men. In older adults, the prevalence of co-morbidities increases[43]. Hypertension, lipid metabolism disorders and chronic low back pain was found to be the most prevalent triad in the elderly[44]. Also, in a Japanese national survey, a trend for a higher point prevalence of LBP complaints for women aged over 65 was reported[45]. It is noteworthy that in 2010, the population of the province of Quebec ranked among the oldest worldwide, behind Japan[45, 46]. Therefore, this increase of recurrent consultations for LBP might become more frequent in aging populations, especially for older women.

Our study confirmed that among the patients with claims-based recurrent LBP, 65% were diagnosed with non-specific LBP with respect to the classification proposed by Cherkin and colleagues[23]. This proportion is lower than the previously reported 84%[21], 76%[47] and 86%[48] for non-specific diagnoses from claims-related databases in the US, but higher than the 53% reported by Martin and colleagues based on a survey[12]. In primary care, other studies have reported that 80% of LBP diagnoses are labeled as non-specific and that no anatomopathological markers helped to explain the patient’s pain[49]. The frequency that we report here may, however, underestimate the true proportion of non-specific LBP diagnoses in the population since our cohort selection was based on the occurrence of 3 diagnoses within a 365-day period, in comparison to single LBP episodes. For instance, it was shown that a group of patients consulting physicians 6 times or more for LBP had fewer non-specific diagnoses than those consulting once (57% vs 84%, respectively)[21]. There are several hypotheses that can explain this over-representation of the non-specific label. First, in studies analyzing administrative databases, we have to keep in mind that physicians or their administrative staff may be using the most common ICD-9 codes in their specialty practice, among which are the non-specific labels. There have also been reports that among physicians, 63% felt that they lacked adequate training in chronic pain management[50]. Between 2000 and 2007, there was a substantial increase in spinal stenosis diagnoses in our analyses. Yet, a study reported that less than 50% of general practitioners were “very confident” in diagnosing specific conditions such as lumbar spinal stenosis and scoliosis in elderly populations. While diagnostic procedures are consistently focused on first identifying “red flags” or suspected serious pathologies, imaging is only recommended after 4 to 7 weeks, if no improvement in the condition is observed (Koes 2010). Therefore, patients diagnosed with a non-specific condition and initially no apparent serious pathology might have successive medical visits with a non-specific label before an exhaustive examination is conducted.

Our study has several limitations related to secondary analyses of administrative health care databases. First, the recent definitions for acute or recurrent LBP could not be used in the context of our study[15, 16]. Indeed, the absence of clinical information on the duration and the intensity of LBP episodes in the administrative data is incompatible with the most recent definitions. Administrative data do not differentiate between a new LBP episode and a claim related to a previous episode that is ongoing[19]. The concept of recurrence in our study is therefore based only on repeatedly consulting a physician. Second, we proposed that three or more diagnoses in a period of one year would identify most of the cases of recurrent claims. A potential information bias thus remains if LBP was a secondary condition for patients presenting with multi-morbidities[51]. Indeed, physicians can only bill one diagnosis claim related to their patient’s condition, a peculiarity of the RAMQ claims database[52]. Selecting patients with only one or two LBP diagnoses could have also allowed for a greater sensitivity, but would have led to less specificity. Ultimately, a future study linking administrative and medical data could optimize both sensitivity and specificity. Time gaps between consultations have not been taken into account in this study. Time gaps between claims have been proposed as an option for improving the accuracy of the identification of the recurrence of a LBP episode. To date, the most frequently used clinical variable to report recurrence is the duration of the event[11]. The mean gap between two distinct LBP episodes was previously reported to be 40 days[19]. Algorithms involving the notion of time gaps between claims have to be further developed to obtain a more specific selection of patients in administrative databases. Medical data linkage to administrative data would also be helpful to increase the validity of such algorithms. Finally, we could not use exactly the same diagnosis categories reported by Vogt or Cherkin and colleagues[23, 47], which decreased the specificity of our results. We only had access to 26 ICD-9 codes with a 4-digit precision. Fortunately, we were granted access to the most frequent LBP diagnoses reported in the literature.