We hypothesized that a new diagnosis of cancer, an exacerbation of chronic pulmonary disease or depression would temporarily shift clinical priorities for both patients and clinicians, leading to changes in the intensity of blood glucose, systolic blood pressure and LDL cholesterol control among persons with underlying diabetes. Our results did not confirm this hypothesis. Based on our analysis, on a population level, none of these disease-specific outcomes was affected by the additional disease burden presented by the new comorbidities. This was suggested by the initial population-level analysis (Fig. ) and re-emphasized by the results of the latent class trajectory analysis (Fig. ). The latent class analyses suggest that over 80% of each subgroup had stable A1c levels over the time frame studied.
Previous investigations have described negative associations between these specific comorbidities and quality of care for DM. For example, increased depression severity has been associated with decreased adherence to DM care recommendations and DM control31,32
, poor glycemic control has been associated with lower levels of lung function33,34
, and cancer survivors may be less likely to receive certain aspects of recommended care for diabetes 6,35
. Other studies suggest that treatment intensification in the face of competing demands may be partially a function of the extent of patients’ symptoms (e.g., chronic pain) or total number of comorbidities.36–38
In this longitudinal analysis of an historical cohort, we did not observe associations between the incident comorbidities and DM 2. There are several possible factors that may contribute to these findings: It is possible that these outcomes are largely a function of well-established self-care behaviors, including dietary habits, exercise patterns and medication adherence. It is also possible that other biopsychosocial factors such as duration of specific conditions, disease burden, depression, personal stress, financial constraints, physical functioning, self-efficacy and social support may influence these outcomes.39–43
Alternatively, clinicians (rather than patients) may perceive “competing demands” from a new diagnosis. For example, clinicians addressing increasing numbers of patient concerns during office visits are less likely to change medication management for diabetes, independent of A1c levels.44
Finally, incident comorbidities in themselves may simply have no bearing on the intensiveness of care that affects these specific process-of-care outcomes in these cohorts. Further investigation of the characteristics of subpopulations characterized by the different trajectories of glycemic control (and other outcomes) will be necessary to explain the processes underlying our population-level results.
In all three subcohorts, LDL and SBP declined overall throughout the time period and were not affected by the incident comorbidities. This may partially reflect secular trends that resulted from changes in clinical recommendations for these parameters for persons with DM and/or the effect of population management initiatives within the organization. (Although these effects would be somewhat attenuated by the varying time periods over which the follow-up measurements occurred.) To evaluate the possibility of underlying secular trends for these outcomes, we assessed average decline in SBP and LDL in the entire cohort of persons with DM (n
27,432) and found that, on average, SBP decreased by 0.75 mmHg per year and LDL by 4.20 mg/dl per year.
We studied values for A1c, LDL and SBP rather than the frequency of measurement of these parameters. Previous evaluations of quality measures in complex patients have concluded that individuals with higher morbidity are more likely to have measurements of these intermediate outcomes and that this in turn is partly (but not completely) mediated by the increased number of visits to clinicians.12,45
Our results, in which none of three measures of patient-clinician contact (continuity of care, primary care visits or specialty care visits) meaningfully affected values of A1c, suggest that quality measures based on frequency of measurement may not reflect the actual values of these health outcomes in persons with multiple morbidities and calls into question measurement frequency as a relevant quality measure.
Our results should be interpreted in the context of several limitations. Administrative and electronic clinical data do not allow full exploration of the myriad biopsychosocial factors that may influence decision-making by both patients and clinicians. Therefore, we were unable to explore potential mechanisms that might explain our findings. For example, in communities of lower socioeconomic status (SES), both patients and clinicians experience increased demands within the clinical encounter, and we did not explore outcomes based on strata of SES.46
In addition, the first clinical diagnosis of an incident comorbidity may not truly represent the first occurrence of the condition, but merely the first time it is brought to clinical attention. In this case, the effect would be to bias the results towards the null (the effect of the ‘incidence’ would be reduced). We attempted to compensate for this limitation by creating a 6-month window of time after the index date during which each cohort member was enrolled, but did not meet criteria for diagnosis of the incident comorbidity. It is also possible that tools or criteria available for diagnosis of a comorbid condition may have changed over the time period of the longitudinal cohort. However, we are unaware of any specific changes in criteria for diagnoses of the comorbid conditions. Criteria for our primary outcome measure of A1c have always been standardized to the Diabetes Control and Complications Trial (DCCT), and internal assays have been stable over time.47
Finally, our results reflect the clinician and patient behavior of members of an integrated health care system that can readily track health outcomes, provide care reminders, and support chronic illness self-management independent of primary care or specialty visits. Given known trends in the organization, it is possible that cohort members may have received independent assistance with lipid management, but that blood pressure management and glycemic control were likely a function of primary care encounters.