Table 3

Association of Physician Continuity With Time to Death or Urgent Admission

Base Analysis—Continuity in Month 1 Captured as Time-Dependent Variable, All Patients Included
Observation TimeContinuity in Month 1Proportion With EventEvents Per 100 PYsUnadjusted HR (95% CI)Adjusted HR (95% CI)
3 monthsNo visits49.3%2421.00 (referent)1.00 (referent)
All visits with unfamiliar physician(s)30.4%1600.96 (0.86–1.08)1.09 (0.97–1.22)
≥1 visit with familiar physician(s)25.2%1320.84 (0.78–0.91)0.87 (0.81–0.95)
6 monthsNo visits58.1%1831.00 (referent)1.00 (referent)
All visits with unfamiliar physician(s)40.2%1150.92 (0.83–1.01)1.04 (0.94–1.15)
≥1 visit with familiar physician(s)37.3%1040.87 (0.81–0.92)0.89 (0.83–0.96)
12 monthsNo visits68.1%1361.00 (referent)1.00 (referent)
All visits with unfamiliar physician(s)51.7%830.88 (0.80–0.96)1.00 (0.91–1.10)
≥1 visit with familiar physician(s)51.8%810.89 (0.84–0.94)0.91 (0.85–0.97)
Sensitivity Analysis—Continuity Captured as Time-Dependent Variable Throughout Entire Duration of Observation, Excluding Patients Without Outpatient Visits
Observation TimeContinuity Throughout Observation TimeProportion With EventEvents Per 100 PYsUnadjusted HR (95% CI)Adjusted HR (95% CI)
3 monthsOutpatient visits but none with familiar physician31.3%1611.00 (referent)1.00 (referent)
≥1 visit with familiar physician(s)0.79 (0.71–0.89)
<50% of visits with familiar physician(s)15.1%1201.00 (0.81–1.23)0.80 (0.64–0.99)
50%–99% of visits with familiar physician(s)15.6%1130.91 (0.80–1.04)0.72 (0.62–0.84)
All visits with familiar physician(s)32.2%1470.90 (0.80–1.00)0.81 (0.72–0.92)
6 monthsOutpatient visits but none with familiar physician41.8%1161.00 (referent)1.00 (referent)
≥1 visit with familiar physician(s)0.86 (0.77–0.95)
<50% of visits with familiar physician(s)22.6%820.96 (0.82–1.12)0.81 (0.69–0.95)
50%–99% of visits with familiar physician(s)25.6%880.997 (0.90–1.11)0.81 (0.72–0.92)
All visits with familiar physician(s)52.4%1210.98 (0.89–1.08)0.88 (0.79–0.98)
12 monthsOutpatient visits but none with familiar physician54.9%861.00 (referent)1.00 (referent)
≥1 visit with familiar physician(s)0.87 (0.80–0.96)
<50% of visits with familiar physician(s)34.9%631.00 (0.89–1.14)0.85 (0.74–0.96)
50%–99% of visits with familiar physician(s)39.9%691.06 (0.96–1.16)0.85 (0.76–0.94)
All visits with familiar physician(s)75.1%1011.01 (0.93–1.11)0.89 (0.81–0.99)

CI = confidence interval; HR = hazard ratio; other abbreviations as in Tables 1 and 2.

  • Covariates included in adjusted models at each time frame include: age, gender, rural versus urban residence, Charlson Comorbidity Index at baseline, number of ED visits for any cause in 6 months before the index HF ED visit, number of outpatient office visits for any cause in the year before the index HF ED visit, specialist involvement during index HF ED visit, outpatient visits for any cause with specialist before index HF ED visit, long-term care facility admission, and presence/absence of diabetes, atrial fibrillation, dementia, and previous MI or coronary revascularization. For the sensitivity analysis, the number of outpatient visits after ED visit were included up to the various time frames.

  • Note that the crude data (proportion with event and events per 100 patient-years [PYs]) are misleading in the sensitivity analysis. For the time-varying groups used in the Cox models, each patient will always start with a usual provider of continuity (UPC) of 1 or 0 after their first follow-up visit. Subsequent follow-up time (i.e., the denominator for the PYs column) will continue to accrue in their starting group if their UPC remains 0 or 1, switch to one of the intermediate groups (0 < UPC <0.5 or 0.5 < UPC <1) as soon as they see a different physician, or will be censored if they have an event. Thus, patients having an event while in the UPC 1 or UPC 0 group will count in the numerator for that group but without further accrual of follow-up time in the denominator; on the other hand, as patients who have not had an event transition into the intermediate groups (0 < UPC <0.5 or 0.5 < UPC <1), their follow-up PYs will accrue in the intermediate group denominators, thus providing misleadingly low “events per 100 PYs” in those groups. This is not an issue in the Cox models used to generate the HR for the sensitivity analysis since the risk sets are re-evaluated at each time point based on the time-varying UPC values.