Abstract
Objective To determine the extent of chronic obstructive pulmonary disease (COPD) hospitalization in easily identifiable high-risk subgroups within a typical primary care practice.
Design Prospective cohort analysis of administrative claims data.
Setting British Columbia.
Participants British Columbia residents who were 50 years or older on December 31, 2014, and received a physician diagnosis of COPD between 1996 and 2014.
Main outcome measures Rate of acute exacerbation of COPD (AECOPD) or pneumonia hospitalization in 2015, broken down by risk identifiers including previous AECOPD admission, 2 or more community respirologist consultations, nursing home residence, or none of these.
Results Of the 242,509 identified COPD patients (12.9% of British Columbia residents ≥50 years), 2.8% were hospitalized for AECOPD in 2015 (0.038 AECOPD hospitalizations per patient-year). The 12.0% with prior AECOPD hospitalization accounted for 57.7% of new AECOPD hospitalizations (0.183 hospitalizations per patient-year); the 7.7% with respirologist involvement accounted for 20.4% (0.102 hospitalizations per patient-year); and the 2.2% in nursing homes accounted for 3.6% (0.061 hospitalizations per patient-year). Those with any of the 3 risk identifiers accounted for only 1.5% more COPD hospitalizations (59.2%) than those with prior AECOPD hospitalization, suggesting prior AECOPD hospitalization is the most important indication of risk. A typical primary care practice held a median of 23 (interquartile range=4 to 65) COPD patients, of whom roughly 20 (86.4%) had none of these risk identifiers. This low-risk majority had only 0.018 AECOPD hospitalizations per patient-year.
Conclusion Most AECOPD hospitalizations occur in patients with previous such admissions. When time and resources are limited, COPD initiatives targeting primary care practices should focus more on the 2 to 3 patients with prior AECOPD hospitalization or more symptomatic disease, and less on the low-risk majority.
Chronic obstructive pulmonary disease (COPD) is a major driver of hospitalization among older adults. Nevertheless, various interventions have been shown to reduce hospitalization for acute exacerbation of COPD (AECOPD), including pulmonary rehabilitation (combination of exercise, education, and behavioural interventions), inhaled β2-agonists, inhaled antimuscarinics, and action plans for self-management of exacerbations that often include oral steroids and oral antibiotics.1-4 In a previous meta-analysis, an integrated multidisciplinary and multitreatment COPD intervention reduced AECOPD hospitalization by 26% compared with usual care.5 Although such a comprehensive intervention has the potential to be clinically important, whether it represents good use of time and resources to patients, providers, and payers depends on the underlying risk of AECOPD hospitalization. This is especially relevant given that participants in COPD therapeutic trials are a poor representation of community COPD patients,6 and that COPD therapeutics have yet to convincingly demonstrate a reduction in mortality or a delay in disease progression.7
Hospitalization due to AECOPD concentrates in patients who are more symptomatic.8 While randomized trials regularly recruit such higher risk patients out of specialty care settings, primary care populations are heterogeneous and contain patients at both high and low risk of COPD exacerbation.
In this retrospective cohort study, we sought to determine the rate of AECOPD hospitalization in primary care patients according to the presence or absence of easily identifiable indicators of higher risk. To make these rates relatable, we used them to estimate the number of subgroup-specific AECOPD hospitalizations that an effective (26% reduction) COPD intervention could prevent, per decade, within a typical family practice. This is relevant given that COPD initiatives aside from smoking cessation and pneumococcal or influenza vaccination are often promoted to primary care providers (PCPs) as initiatives they should consider for all of their COPD patients, even those with mild symptoms.9-12
METHODS
Source data
British Columbia maintains, with minimal under-reporting and misclassification,13 linkable databases on all physician services, hospitalizations, and dispensed prescription medication for all individuals in its publicly funded health care system.14 These 3 databases, along with Medical Services Plan registration data and vital statistics data for date of death, were linked using encrypted patient identifiers for the period of January 1, 1996, to December 31, 2015, and composed our source data.15-19 Thus, our source population included the roughly 4.5 million residents of British Columbia.14*
Setting
Residents of British Columbia do not pay directly for health care services (ie, there are no user fees) and most residents (94% of Canadians older than 65 years)20 have a family physician PCP. Specialist services can be accessed, when needed, by referral from the PCP or during acute care hospitalization. Although health care provider services are publicly funded, medication costs are typically borne by the individual, with some or all medication costs being publicly funded for senior citizens and for some individuals with challenging health or socioeconomic circumstances.
Study population
Participants were included if they were alive and resided in British Columbia on December 31, 2014; received a community or hospital physician claims diagnosis of COPD (ICD-9 codes21 491, 492, or 496; ICD-10 codes22 J41 to J44) between January 1, 1996, and December 31, 2014 (note that this includes both incident and prevalent cases); and were 50 years or older on December 31, 2014.
We required only a single claims diagnosis of COPD as this is a validated (85.0% sensitivity, 78.4% specificity),23 established, and recommended method of detecting COPD in administrative claims data.23-25 Our cohort hence consisted of all British Columbia residents with 1 or more community or hospital claims diagnoses for COPD from 1996 to 2014 who were alive, 50 years of age or older, and registered for health services on December 31, 2014.
Detecting AECOPD-related hospitalization
Pneumonia and AECOPD can be difficult to distinguish at times. To avoid underestimating the hospitalization rate, we considered AECOPD-related hospitalization to occur whenever the primary discharge diagnosis was COPD (ICD-9 codes 491, 492, or 496; ICD-10 codes J41 to J44), unspecified lower respiratory tract infection (ICD-9 code 519.8, ICD-10 code J22), or pneumonia (ICD-9 codes 480 to 486, ICD-10 codes J12 to J18).
Outcomes
Primary outcomes included the rate and proportion of all COPD patients hospitalized for AECOPD annually (assessed during 2015). When broken down by subgroups, this consisted of prior AECOPD hospitalization, 2 or more community respirologist consultations (indicating patients with more symptoms), nursing home residence, and none of the aforementioned risk indicators.
Secondary outcomes included the number of COPD patients in a “typical” primary care practice (ie, refers to a population median and makes no assumptions about practice style or practice demographic characteristics) and the proportion of COPD patients who fall within each of the identified risk categories; and the estimated number of AECOPD hospitalizations that could be prevented over time by 1 PCP applying an effective intervention that lowers AECOPD hospitalization by 26% (this being the relative risk reduction demonstrated in a systematic review of an integrated multidisciplinary and multitreatment COPD intervention versus usual care5).
Statistical analysis
Outcomes were determined during the 1-year period after cohort construction (January 1, 2015, to December 31, 2015). In determining the number of COPD patients per PCP practice, we considered all PCPs with at least 1 billing in 2015 to be eligible (ie, all PCPs who were still practising at the end of 2014) and assigned each COPD patient to the eligible PCP with whom they had the most contact during the cohort period (with a tie going to the PCP with most recent contact). The analysis consisted of a simple determination of rates and summary statistics for the identified subgroups.
RESULTS
Construction of the cohort is outlined in Figure 1. There were 242,509 adults 50 years of age or older who met our COPD cohort definition and who were alive on December 31, 2014, representing 12.9% of the 1.9 million British Columbia residents 50 years or older in 2014.26 Low-risk patients were represented equally by male and female patients and had a mean (SD) age of 68.2 (10.8) years. Individuals in the higher-risk categories were older and more were female. Detailed baseline characteristics are provided in Table 1. The use of COPD therapeutics was higher in those with indications of higher risk.
Flow diagram of cohort construction
Baseline characteristics of COPD cohort
A typical PCP looked after a median of 23 COPD patients (interquartile range=4 to 65). Hospitalization rates of AECOPD in the different risk categories are shown in Table 2. The 12.0% with prior AECOPD hospitalization (2 to 3 patients per PCP) accounted for 57.7% of new AECOPD hospitalizations, at a rate of 0.183 AECOPD hospitalizations per patient-year. In contrast, roughly 20 patients (86.4%) were community dwelling, had no prior AECOPD hospitalization or respirologist involvement, and experienced only 0.018 AECOPD hospitalizations per patient-year, which is roughly one-tenth as often as those with prior admission.
AECOPD hospitalizations in 2015
Rates of AECOPD hospitalization for those with respirologist involvement (0.102 per patient-year) and those with nursing home residence (0.061 per patient-year) were midway between these 2 extremes. However, the proportion of AECOPD hospitalizations attributable to those with any of our 3 risk indicators was only marginally higher (59.2%) than the attributable proportion determined by considering only those with previous AECOPD hospitalization (57.7%).
DISCUSSION
A single PCP with 23 COPD patients could expect to observe 8.7 AECOPD hospitalizations per decade among their own COPD patients. Of these, the PCP could prevent 2.3 AECOPD hospitalizations per decade by applying a comprehensive COPD intervention to all COPD patients within their practice. More than half (57.7%) of the total avoided hospitalizations would come from focusing on the 2 to 3 patients (12.0%) with previous hospital admission. Although 40.8% of total AECOPD hospitalizations would still occur in the 86.4% of patients at low risk, applying an effective COPD intervention to all low-risk COPD patients within a PCP practice would prevent only 3.7 AECOPD hospitalizations over a 40-year practice career.
Our 12.9% COPD prevalence for British Columbia residents 50 years or older at time of diagnosis is midway between the Vancouver cohort in the BOLD (Burden of Obstructive Lung Disease) study—which found that 11.1% of Canadians 40 years or older responding to random digit dialing had physiologically defined COPD (as indicated by spirometry test)27—and the 15.5% COPD prevalence for Canadians 50 or older found using a similar administrative claims definition in the province of Ontario.24 Our hospitalization rate in those with prior AECOPD admission (0.183 per patient-year) is comparable to control hospitalization rates observed in large COPD trials such as TORCH (Towards a Revolution in COPD Health) (0.19 per patient-year)28 and UPLIFT (Understanding Potential Long-Term Impacts on Function with Tiotropium) (0.16 per patient-year).29 Our observation that hospitalization concentrates in higher-risk patients aligns with both our clinical experience and the POET-COPD (Prevention of Exacerbations with Tiotropium in COPD) trial that reported 13.6% of COPD patients with frequent exacerbations at baseline accounted for 56.6% of exacerbation-related hospitalizations.8
Strengths and limitations
Although population data improve generalizability, administrative claims data are not designed for research purposes. Identifying COPD patients using a single physician claims diagnosis is reported to have 85.0% sensitivity and 78.4% specificity when compared with the criterion standard of an expert panel reviewing clinical records.22 While we could have increased our specificity by requiring 2 COPD diagnoses or by limiting our cohort to those who actively use COPD medications, this would only have reduced the total number of preventable AECOPD admissions per PCP practice by reducing the size of the COPD cohort. The same would be true if we limited AECOPD hospitalization to those with a clear COPD exacerbation by excluding diagnoses for pneumonia or respiratory infection that were not specifically associated with COPD. A clearer AECOPD hospitalization diagnosis would only reduce the number of hospitalizations, and hence simultaneously reduce the number of hospitalizations that could be prevented. Our assumption of a 26% reduction in AECOPD hospitalization also derives from studies enrolling COPD patients at high risk of exacerbation.5 It is unclear that a relative risk reduction this large would transfer to those at lower risk.10,30-32 Hence, we may be overestimating the number of AECOPD admissions that an effective COPD intervention program could prevent.
Although our risk categories are not traditional (previous AECOPD hospitalization, respirologist involvement, and nursing home residence), they are also a strength given providers and payers can easily identify and target these patient groups if desired. Health authorities could, for instance, reach out to PCPs, advise them of their COPD patients who appear at highest risk of AECOPD hospitalization based on prior such admission, and offer to make a respiratory rehabilitation program available to those individuals.33
Conclusion
Primary care COPD patients vary in risk of AECOPD hospitalization. Of the median 23 COPD patients in a typical practice, the 2 to 3 (12.0%) with prior AECOPD hospitalization will be responsible for 57.7% of new COPD hospitalizations in the following year. Because health systems have limited resources, community-based programs intent on reducing AECOPD hospitalization should target the high-risk minority as opposed to the low-risk majority.
Notes
Editor’s key points
▸ To prevent 1 acute exacerbation of chronic obstructive pulmonary disease (AECOPD) hospitalization annually, a multitreatment COPD intervention that reduces hospitalization by 26% (estimated relative risk reduction from meta-analysis) would need to be offered to 213 low-risk patients compared with 21 patients with previous AECOPD hospitalization.
▸ For a typical practice, applying such a COPD intervention to the low-risk majority (86.4% of COPD patients) would prevent 0.9 AECOPD hospitalizations per practice per decade (ie, prevent only 3.7 AECOPD hospitalizations over a 40-year practice career).
▸ Primary care providers should focus COPD interventions (aside from smoking cessation and pneumococcal or influenza vaccination) on their few patients with previous AECOPD hospitalization or more symptomatic disease.
Points de repère du rédacteur
▸ Pour prévenir 1 hospitalisation due à une exacerbation aiguë d’une maladie pulmonaire obstructive chronique (EAMPOC) par année, une intervention à traitements multiples pour la MPOC, qui réduit de 26 % les hospitalisations (réduction du risque relatif estimé selon une méta-analyse), devrait être offerte à 213 patients à faible risque par rapport à 21 patients ayant antérieurement été hospitalisés pour une EAMPOC.
▸ Dans une pratique typique, la mise en œuvre d’une telle intervention pour la MPOC chez la majorité des patients à faible risque (86,4 % des patients atteints de MPOC) préviendrait 0,9 hospitalisation due à une EAMPOC par pratique par décennie (p. ex. la prévention de seulement 3,7 hospitalisations pour EAMPOC au cours d’une carrière de 40 ans).
▸ Les professionnels des soins primaires devraient diriger plutôt les interventions pour la MPOC (exception faite de la cessation du tabagisme et de la vaccination contre la grippe) vers leurs quelques patients antérieurement hospitalisés en raison d’une EAMPOC ou souffrant d’une maladie plus symptomatique.
Footnotes
↵* All inferences, opinions, and conclusions drawn in this research are those of the authors and do not reflect the opinions or policies of the Data Steward (Population Data BC).
Contributors
All authors contributed to the concept and design of the study; data gathering, analysis, and interpretation; and preparing the manuscript for submission.
Competing interests
None declared
This article has been peer reviewed.
Cet article a fait l’objet d’une révision par des pairs.
- Copyright © 2023 the College of Family Physicians of Canada