Session Title: Epidemiology and Health Services II & III
Session Type: Abstract Submissions (ACR)
Background/Purpose: Hospital administrative datasets offer an opportunity to study the relatively rare occurrence of auto-immune rheumatic diseases (AIRD). Under-coding is a known limitation however. “Look-back” periods can be used to compensate for under-coding within the (“index”) admission of interest by identifying AIRD codes in preceding admissions for each patient. However if the timespan of the dataset is fixed, increasing the look-back period to identify more AIRD patients reduces the duration of the primary analysis period (Figure 1). This study aimed to examine the effect of different look-back periods for identifying AIRD patients in an evaluation of the relationship between AIRD status and mortality following myocardial infarction (MI).
Methods: This study utilises a population-based hospital admission dataset with data available from 1 July 1998 to 30 June 2007. MI and AIRD status were identified using relevant International Classification of Diseases (ICD) codes. Six scenarios were defined for ascertainment of AIRD status (A to F: Figure 1), ranging from no look-back to a 5 year look-back period. Thirty-day and 1-year mortality rates were calculated from the date of the index MI. A logistic regression model was fitted with mortality as the outcome, AIRD status as the exposure and adjustment for relevant covariates. We compared the relationship between AIRD status and post-MI mortality for each scenario.
Results: As the duration of the look-back period increased (progressing from scenario A to F), the prevalence of AIRD increased from 0.7% (n=632 of 86,841 patients: scenario A) to 2.2% (n=998 of 45,447 patients: scenario F). The number of patients identified with MI decreased from scenario A to F due to the progressive reduction in the duration of the primary analysis period. Adjusted odds ratios for all-cause mortality for the AIRD group are shown in Figure 2. When no look-back period was used, no significant relationship between AIRD and 30-day mortality was identified (OR = 1.13, 95%CI 0.90-1.42). As the look-back period increased, a statistically significant relationship was obtained with an increase in the odds ratio, before stabilising with a look-back period of 3 years or longer in duration.
Conclusion: Using a look-back period to identify AIRD status changed the significance of our findings. Based on the dataset in question, there is no advantage of extending the look-back period beyond a duration of three years, with an analysis period of six years.
Figure 1 –Index MI ascertainment and look-back period scenarios
Figure 2 –Adjusted Odds Ratio for 30-day and 12-month mortality for AIRD patients
S. Van Doornum,
« Back to 2013 ACR/ARHP Annual Meeting
ACR Meeting Abstracts - https://acrabstracts.org/abstract/methodology-of-determining-an-appropriate-look-back-period-to-identify-autoimmune-rheumatic-disease-in-a-study-of-post-myocardial-infarction-mortality/