Document resource
WHAT IS ALREADY KNOWN ON THIS TOPIC Hemoglobin A1c (A1c) treatment goals often focus on recent A1c measurements without considering the cadence and stability of A1c over time. In older adults with diabetes, A1c stability within patient-specific target ranges with upper and lower bounds may inform risk prediction for major adverse events.WHAT THIS STUDY ADDS Among older adults with diabetes (65 years or older) from the Veterans Health Administration and Kaiser Permanente, lower A1c time in range was associated with higher mortality. Instrumental variable models that controlled for unmeasured factors showed that lower A1c time in range was associated with increased mortality and cardiovascular outcomes.HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY Greater A1c stability within patient-specific ranges is associated with fewer deaths and cardiovascular outcomes in older adults with diabetes.Introduction Associations among chronic hyperglycemia, mortality, and diabetes complications are complex. Higher average hemoglobin A1c (A1c) is associated with microvascular (eg, kidney and eye) complications. 1 2 However, there is a non-linear and perhaps a U-shaped relationship between average A1c and mortality.3 Treating patients with new-onset diabetes to lower A1c levels may accrue benefits over time,4 but targeting near-normal A1c levels in those with established type 2 diabetes does not reduce cardiovascular disease (CVD) or mortality.2 5 6 In addition, increased A1c variability is associated with higher risks of mortality, CVD, kidney disease, and retinopathy.7–10Diabetes clinical practice guidelines account for these complexities by recommending different A1c treatment goals based on life expectancy, comorbidities, and complications.11–14 Nonetheless, clinical treatment strategies often focus on reducing elevated A1c levels without considering the risks associated with increasing A1c variability or whether treatment goals should include a lower threshold. This approach may lead to potential overtreatment.15To address the need for a clinical measure that captures A1c cadence and stability, we developed a measure of A1c time-in-range (A1c TIR).16 A1c TIR represents the percentage of time over a 3-year period that A1c levels are within patient-specific and guideline-directed ranges.12 In an initial cohort of veterans, we showed that higher A1c TIR is associated with lower risk of mortality and several diabetes complications.16–18The generalizability and utility of A1c TIR as a risk predictor depend on its confirmation in other patient cohorts and stronger evidence of potential causal associations with major adverse outcomes. The objectives of this study are to examine A1c TIR as risk predictors for mortality and diabetes complications in different patient cohorts and apply methods that account for unmeasured confounding. We studied patients with diabetes from two large healthcare systems that included two Kaiser Permanente (KP) health systems and an updated cohort from the Veterans Health Administration (VA).Research design and methods Study population This was a retrospective cohort study with patients followed longitudinally. We used electronic health record and associated administrative data from January 1, 2004, through December 31, 2018. The two study cohorts were comprised of patients 65 years or older at the start of enrollment, including VA patients who were dually enrolled in Medicare and KP patients drawn from KP Northwest and KP Mid-Atlantic states.For eligibility criteria, we defined a diabetes diagnosis as having at least two outpatient or one inpatient diagnosis code (International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM): 250.XX, 357.2, 362.0X, 366.41 and International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM): E10.X, E11.X, E13.X, where X indicates all subcodes) or a prescription for a diabetes medication. This includes patients with type 1 and type 2 diabetes. Patients were excluded if they had less than four A1c tests or if A1c tests were more than 12 months apart during a 3-year baseline that occurred between January 1, 2005, and December 31, 2017. Patients were removed from the cardiovascular and microvascular outcome analyses, respectively, if they had a prevalent diagnosis for one of these outcomes during baseline (online supplemental figures S1 and S2).SP110.1136/bmjdrc-2025-005188.supp1Supplementary dataStudy variables Patient-level A1c TIR Applying previous methods, 16–18 A1c TIR was calculated as the percentage of days during a 3-year baseline that a patient had A1c levels within their individualized target range. To obtain a patient’s A1c target range, we applied the VA/Department of Defense Diabetes Clinical Practice Guidelines12 which proposes four different A1c ranges, 6.0–7.0%, 7.0–8.0%, 7.5–8.5%, and 8.0–9.0%, based on a patient’s life expectancy and the presence and severity of diabetes complications. We estimated life expectancy using a predictive model of clinical and administrative data and separated patients into groups of less than 5 years, 5–10 years, and 10 or more years.19 The Diabetes Complications Severity Index (DCSI)20 was used to identify diabetes complications and their severity. The DCSI is a validated index that scores diabetes complications in seven categories using ICD-9/10 codes and laboratory tests, with a range between 0 and 13. We assessed life expectancy and prevalent diabetes complications for each patient in a 1-year period preceding the baseline to set the initial A1c target range. A1c target ranges and time in range were updated annually during the 3-year baseline to adjust for comorbidities or new diagnoses that impact mortality or severity of complications. A1c TIR was calculated using linear interpolation between A1c values and lab test dates. The total days that A1c values were within the corresponding target range were summed and divided by the total days of baseline (1095 days) and expressed as a percentage. A1c TIR was divided into five categories: 0% to <20%, 20% to <40%, 40% to <60%, 60% to <80%, and 80% to 100% (reference group).We also studied the direction of out-of-range levels by grouping patients into four mutually exclusive categories based on the percentage of days that A1c was within, below, or above their unique target range. The groups were ≥60% A1c TIR (reference group), ≥60% A1c time below range (TBR), ≥60% A1c time above range (TAR), and a mixed group comprised of patients whose A1c TIR, TBR, and TAR were all below 60%. A≥60% threshold was used because it comprised a majority of time spent in the respective category. For each patient, the sum of A1c TIR, TBR, and TAR equaled 100%.Clinician A1c TIR as an instrumental variable (IV) We employed an IV to limit bias from unmeasured factors or confounding in the association between patient-level A1c TIR and outcomes. 21 The IV approach helps mimic randomization to allow for less biased estimates. A suitable IV is one that has a strong influence on the likelihood of receiving a treatment but has no direct effect on the outcome and is not associated with confounders.22 Clinical practice variation is commonly used as an IV.23–27 We used clinician A1c TIR as an IV by calculating patient-level A1c TIR at the clinician level. For analysis, patients were aligned to the specific clinician who ordered the most A1c tests for them during the baseline. We then calculated A1c TIR for all the patients assigned to each clinician during baseline and averaged these values. If a clinician had fewer than 10 assigned patients, we used the A1c TIR for all clinicians at the treating site for VA patients and by health system for KP patients.Covariates Baseline covariates included age, sex (self-reported), race/ethnicity (self-reported), DCSI Score, comorbidities, medications, laboratory tests, clinician type, calendar quarter in which a patient entered the outcome period, and VA Medical Center or KP health system ( online supplemental table S1). The Elixhauser Comorbidity Index28 was used to measure comorbidities during the baseline. Diabetes medications dispensed during baseline were categorized by class. Medication adherence was defined as 80% or greater days with on-hand supply by calculating the medication possession ratio (ie, days’ supply/total number of days) for all prescribed diabetes medications. Laboratory measures included baseline serum creatinine, serum albumin, urine albumin-to-creatinine ratio, and blood lipids (ie, high-density lipoprotein cholesterol [HDL], low-density lipoprotein cholesterol [LDL], triglycerides). Clinical measures included average body mass index and blood pressure during baseline. Multiple measures were averaged and then grouped into categories using clinical criteria (eg, low, normal, high) and a separate category was used for missing values. We also measured the number of A1c tests, the patient’s average A1c level, and A1c SD during baseline. Clinician type included physician, nurse practitioner, physician assistant, or other, and if they were a primary care clinician.Outcomes Outcomes were mortality and new cardiovascular and microvascular complications. After excluding patients with prevalent diagnoses during the baseline period, we determined incident cardiovascular outcomes from the DCSI categories for cardiovascular, cerebrovascular, and peripheral vascular diseases and incident microvascular complications from the categories for retinopathy, neuropathy, and nephropathy. To determine all-cause mortality, we used the VA Vital Status File for VA patients, data from the Social Security Administration, state-level vital health status records, and electronic health record information for KP patients.Statistical analysis Descriptive analyses were examined for patient-level characteristics and clinical variables. We estimated the effect of A1c TIR on time to death, cardiovascular outcomes, and microvascular outcomes using adjusted Cox proportional hazards regression models. Results from the Cox models were also assessed for violation of the proportional hazards assumption with Schoenfeld residuals. When the proportional hazard assumption did not hold, the effects of A1c TIR on outcomes were estimated with accelerated failure time models. Follow-up began immediately after the 3-year baseline. Patients were censored on death for the cardiovascular and microvascular outcomes, if they disenrolled from the healthcare system, or at the end of the study period. All adjusted models included the main explanatory variables and all covariates. Models were run separately for VA and KP cohorts.For IV models, we used a two-stage residual inclusion approach. In the first stage, patient-level A1c TIR was regressed on the IV (clinician A1c TIR) and all covariates using a linear regression, applying the criterion that F-statistic less than 10 indicates a weak instrument.29 In the second stage, we included the residual from the first stage as a covariate in a multivariable Cox proportional hazards regression, which was performed separately for mortality, cardiovascular, and microvascular outcomes.We conducted several additional analyses. A key assumption in IV models is that the instrument influences outcomes only through effects on patient-level A1c TIR. We examined this by studying the association of A1c TIR on mortality, cardiovascular, and microvascular outcomes in patients without diabetes based on the assumption that clinician A1c TIR should not influence their outcomes.30 We created a falsification sample of patients with asthma, chronic obstructive pulmonary disorder (COPD), and heart failure but without diabetes. Using the second stage equations, we predicted outcomes using clinician A1c TIR as an IV and all covariates. Finally, because patients are at risk for diabetes complications and mortality at the same time, we used Fine and Gray’s competing risk model for cardiovascular and microvascular outcomes. Analyses were conducted using STATA V.18.Data and resource availability Patient-level data from the Department of Veterans Affairs and KP contain protected health information and cannot be shared. The source code used for generating Elixhauser comorbidities, DCSI index scores, and samples of analytical analyses used for this study will be made available on reasonable request.Results The study groups included 386 287 VA patients and 24 885 KP patients ( table 1). The VA cohort was predominantly male sex (99%) and of white ethnicity (86%) with mean age of 74.3 (SD 5.8) years, mean baseline A1c of 6.9% (SD 1.0) and 79% had prevalent CVD. The KP cohort was 51% female sex and 37% were non-white race/Hispanic ethnicity. KP patients had a mean age of 72.3 (SD 5.7) years, mean baseline A1c of 7.0% (SD 1.0), and 57% had prevalent CVD. Baseline diabetes complications as measured by the DCSI were similar among VA patients (score 2.8; SD 2.2) and KP patients (score 2.9; SD 2.3). Patients were assigned to four A1c target ranges at the end of baseline (online supplemental table S2). During an average follow-up of 5.7 years (SD 3.4), 56% of VA patients died. KP patients had an average follow-up of 6.1 years (SD 3.4) and 39% died.Table 1Baseline characteristics of the study populationsCharacteristicsVA(n=3 86 287)KP(n=24 885)Age, mean (SD) years74.3 (5.8)72.3 (5.7)Sex: male381 144 (98.7%)12 300 (49.4%)Race/ethnicity White333 921 (86.4%)15 599 (62.7%) Black44 477 (10.7%)5351 (21.5%) Asian1377 (0.4%)1445 (5.8%) Other5068 (1.3%)1510 (6.1%) Hispanic4443 (1.2%)980 (3.9%)HbA1c, mean (SD), %6.9 (1.0)7.0 (1.0)HbA1c SD, mean (SD), %0.51 (0.45)0.54 (0.48)Number of A1c tests, mean (SD)5.6 (2.0)6.4 (2.3)LDL cholesterol, mean (SD), mg/dL92.5 (31.9)90.3 (26.4)HDL cholesterol, mean (SD), mg/dL41.3 (20.9)48.9 (14.2)Triglycerides, mean (SD), mg/dL154.2 (85.9)150.3 (82.8)Systolic blood pressure, mean (SD, mm Hg)134.0 (11.6)133.1 (11.7)DCSI Score, mean (SD)2.8 (2.2)2.9 (2.3)BMI, mean (SD)30.2 (5.2)31.0 (6.2)Medications Sulfonylurea227 512 (58.9%)10.932 (43.9%) Metformin156 456 (40.5%)13 425 (54.0%) Insulin71 373 (18.5%)5499 (22.1%) Thiazolidinedione43 660 (11.3%)1942 (7.8%) Other9055 (2.3%)495 (2.0%) Diabetes medication adherence—proportion of days covered≥80%233 286 (60.5%)16 067 (64.6%)Provider type Physician75.1%94.4% Nurse practitioner or physician assistant22.3%2.2% Other2.6%3.5%Health conditions during baseline Diabetes: type 13201 (0.8%)203 (0.8%) Cardiovascular disease304 459 (78.8%)14 215 (57.1%) Heart failure123 298 (31.9%)4715 (19.0%)A1c time in range 80% to 100%64 827 (16.8%)6101 (24.5%) <80%57 169 (14.8%)3973 (16.0%) 40% to <60%50 139 (13.0%)3293 (13.2%) 20% to <40%66 099 (17.1%)4044 (16.3%) 0% to <20%148 052 (38.3%)7474 (30.0%)BMI, body mass index; DCSI, Diabetes Complications Severity Index; HbA1c, hemoglobin A1c; HDL, High-density lipoprotein; KP, Kaiser Permanente; LDL, Low-density lipoprotein; VA, Department of Veterans Affairs.A1c TIR and clinical outcomes in VA patients In adjusted Cox proportional hazards models, lower A1c TIR was associated with all study outcomes in VA patients. When compared with higher A1c TIR (80–100%), lower A1c TIR was associated with a higher risk of death (A1c TIR 0 to <20%, HR, 1.22; 95% CI 1.20 to 1.23) ( table 2). Tests of the proportional hazards assumption showed that the relative risk of mortality was not constant over time. Therefore, we also estimated mortality using accelerated failure time models, which showed similar results (online supplemental table S3). Survival probability curves showed increased mortality associated with lower A1c TIR (figure 1A). Similarly, lower A1c TIR was associated with increased risks of cardiovascular outcomes (A1c TIR 0 to <20%; HR 1.10; 95% CI 1.07 to 1.13) and microvascular outcomes (A1c TIR 0 to <20%; HR 1.03; 95% CI 1.01 to 1.06).Table 2Adjusted Cox regression estimates for the association of A1c time in range and time to mortality and new diabetes complications from fully adjusted and instrumental variable modelsA1c time in rangeMortality(n=386 287)Cardiovascular complications(n=81 828)Microvascular complications(n=70 470)VA patientsAdjusted Cox modelIV modelAdjusted Cox modelIV modelAdjusted Cox modelIV model80% to 100%1.01.01.01.01.01.060% to <80%1.05(1.03, 1.07)1.07(1.03, 1.11)1.05(1.02, 1.07)1.07(1.03, 1.11)1.03(1.00, 1.06)1.06(1.01, 1.10)40% to <60%1.10(1.08, 1.12)1.07(1.03, 1.12)1.03(1.00, 1.07)1.07(1.03, 1.12)1.04(1.01, 1.08)1.08(1.03, 1.13)20% to <40%1.11(1.10, 1.13)1.14(1.09, 1.18)1.06(1.03, 1.09)1.14(1.09, 1.18)1.04(1.01, 1.07)1.10(1.05, 1.15)0% to <20%1.22(1.20, 1.23)1.21(1.19, 1.23)1.10(1.07, 1.13)1.22(1.18, 1.27)1.03(1.01, 1.06)1.17(1.12, 1.21)KP patientsMortality(n=24 885)Cardiovascular complications(n=10 670)Microvascular complications(n=5468)80% to 100%1.01.01.01.01.01.060% to <80%1.20(1.12, 1.30)1.14(1.05, 1.23)0.97(0.90, 1.05)1.15(1.01, 1.31)1.05(1.00, 1.06)1.10(0.90, 1.34)40% to <60%1.27(1.18, 1.38)1.19(1.10, 1.29)1.06(0.97, 1.15)1.28(1.12, 1.47)0.98(0.87, 1.11)1.23(0.99, 1.52)20% to <40%1.23(1.14, 1.33)1.17(1.08, 1.26)0.99(0.91, 1.08)1.19(1.04, 1.37)0.92(0.81, 1.05)1.10(0.89, 1.36)0% to <20%1.36(1.27, 1.45)1.29(1.21, 1.39)1.07(0.99, 1.15)1.36(1.21, 1.53)0.97(0.87, 1.08)1.04(0.86, 1.25)Models included all covariatesIV, instrumental variable (clinician A1c time in range); KP, Kaiser Permanente; VA, Department of Veterans Affairs.Figure 1Kaplan-Meier estimates for time to mortality by hemoglobin A1c time-in-range.Results from IV models were consistent with the adjusted models (table 2). Clinician A1c TIR predicted patient-level A1c TIR in first-stage models of mortality (F=436; p<0.001), cardiovascular outcomes (F=108; p<0.001), and microvascular outcomes (F=94; p<0.001). When compared with higher A1c TIR (80–100%), lower A1c TIR was associated with higher mortality (A1c TIR 0% to <20%; HR, 1.21; 95% CI 1.19 to 1.23), cardiovascular outcomes (A1c TIR 0% to <20%; HR, 1.22; 95% CI 1.18 to 1.27), and microvascular outcomes (A1c TIR 0% to <20%; HR 1.17; 95% CI 1.12 to 1.21).Patients were then grouped into four categories based on the proportion of days during baseline in which A1c was within, above, or below their target range. Percentages of patients in each category were: ≥60% A1c TIR (31.6%), ≥60% A1c TBR (42.9%), ≥60% A1c TAR (9.3%), and the mixed group (16.3%). Greater A1c TBR (≥60%) was associated with increased mortality and cardiovascular outcomes and a lower risk of microvascular outcomes. Greater A1c TAR (≥60%) was associated with increased mortality, cardiovascular, and microvascular outcomes (table 3).Table 3Adjusted HRs (95% CI) predicting mortality and incident diabetes complications by ≥60% time out-of-range categoriesCategoryMortalityCardiovascular outcomesMicrovascular outcomes≥60% time below range* VA patients (n=165 740)1.15 (1.13 to 1.16)1.07 (1.04 to 1.09)0.96 (0.94 to 1.00) KP patients (n=7399)1.26 (1.19 to 1.33)1.07 (0.99 to 1.16)0.97 (0.86 to 1.09)≥ 60% time above range* VA patients (n=35 784)1.11 (1.09 to 1.13)1.09 (1.06 1.12)1.16 (1.12 to 1.19) KP patients (n=3385)1.08 (1.00 to 1.17)1.04 (0.96 1.12)0.93 (0.83 to 1.04)Mixed group* VA patients (n=62 766)1.06 (1.05 to 1.08)1.02 (0.99 to 1.04)1.04 (1.01 to 1.07) KP patients (n=4027)1.12 (1.05 to 1.19)1.06 (0.98 to 1.15)0.97 (0.87 to 1.08)Models included all covariates.*Referent: ≥60% A1c time in range.KP, Kaiser Permanente; VA, Department of Veterans Affairs.A1c TIR and clinical outcomes in KP patients Adjusted Cox proportional hazards models showed that lower A1c TIR was associated with a higher risk of death (A1c TIR 0% to <20%, HR, 1.36; 95% CI 1.27 to 1.45) in KP patients ( table 2). Results were confirmed in accelerated failure time models (online supplemental table S3). Survival probability curves showed increased mortality with lower A1c TIR (figure 1B).In IV models, clinician A1c TIR predicted patient-level A1c TIR in first-stage models of mortality (F=472, p<0.001), cardiovascular outcomes (F=583; p<0.001), and microvascular outcomes (F=269; p<0.001). When compared with higher A1c TIR (80–100%), lower A1c TIR was associated with mortality (A1c TIR 0% to <20%, HR, 1.29; 95% CI 1.21 to 1.39), and cardiovascular outcomes (A1c TIR 0% to <20%, HR, 1.36; 95% CI 1.21 to 1.53).To examine the impact of out-of-range levels, the percentages of patients in each category were: ≥60% A1c TIR (40.5%), ≥60% A1c TBR (29.7%), ≥60% A1c TAR (13.6%), and the mixed group (16.2%). Greater A1c TBR (≥60%) and A1c TAR (≥60%) were each associated with increased mortality (table 3).Additional analyses We identified 284 296 VA patients and 23 172 KP patients without diabetes who had heart failure, COPD, and asthma ( online supplemental table S4). The IV, clinician A1c TIR, did not predict mortality, cardiovascular, or microvascular outcomes in these patients (online supplemental table S5). The frequency and timing of A1c tests to calculate A1c TIR are based on clinical care decisions. Therefore, we performed analyses stratified by the median number of A1c tests and the median of days between A1c tests. Mortality risks were similar to the main results (online supplemental tables S6 and S7). In competing risk models that predicted outcomes with the competing risk of mortality, VA patients had increased risk of both cardiovascular and microvascular outcomes with lower A1c TIR (online supplemental table S8).Discussion We showed that A1c stability within patient-specific target ranges is associated with major adverse outcomes in older adults with diabetes from two large health systems. Lower A1c TIR was associated with a higher risk of death whether out-of-range levels were due to greater time above or below patients’ target ranges. IV models that controlled for unmeasured confounding showed associations between lower A1c TIR and higher risk of mortality and cardiovascular outcomes. The risk of microvascular outcomes was slightly increased among VA patients. There was no association between the IV and outcomes among patients without diabetes, which reduces the likelihood that the IV models were affected by confounders. Together, these results show that lower A1c TIR identifies older adults with diabetes who are at increased risk of major adverse outcomes.Most clinical practice guidelines propose treatment goals with higher A1c targets in older adults with diabetes, comorbidities, and/or limited life expectancy.12–14 These treatment recommendations are often applied prospectively to A1c measurements. However, there is little guidance on the risks associated with A1c stability over time or when levels fall above or below appropriate A1c target ranges. Our findings suggest that A1c stability within patient-specific ranges may be used in glycemic management to identify patients at increased risk of major adverse outcomes. These results also support a role for shared decision-making in older adults when setting and sustaining lower A1c goals that fall outside target ranges.30Results from both adjusted Cox models and IV models were consistent among VA patients, whereas IV models revealed an association between A1c TIR and cardiovascular outcomes in KP patients that was not evident in the adjusted models. This may be due to unmeasured confounding variables that biased the effects in the adjusted models. Examples of such variables may include patients’ approaches to self-care (eg, nutrition and exercise) or clinicians’ cognitive biases or clinical inertia. Also, IV models estimate a local average treatment effect, which is the effect in the subset of patients whose A1c TIR is influenced by their clinician’s A1c TIR. Multivariable Cox proportional hazards models estimate an average treatment effect, which is the effect averaged across all patients, whether or not their A1c TIR is influenced by their clinician’s A1c TIR. A significant local average treatment effect may be washed out when combined with all patients in the population.Glycemic variability can be measured over shorter and longer time periods. A1c TIR focuses on longer-term glucose management. Observational studies show higher risks of mortality and diabetes complications with increased A1c variability,6–10 but there is no generally accepted measure to express A1c stability. Proposed variability measures include A1c SD, coefficient of variation, or absolute change in A1c.7 31 32 However, these variance measures alone fail to consider patient-specific A1c target ranges. A1c TIR integrates both A1c levels and the time spent in individualized ranges and expresses A1c stability in a single measure. Continuous glucose monitoring (CGM) measures intraday and day-to-day readings and similarly expresses variability based on time-in-range metrics. CGM is effective at reducing A1c, hypoglycemia, and hospitalizations,33–35 but data are lacking on its effects on major adverse outcomes. CGM use is growing, but most older adults continue to have glycemic management monitored with A1c levels.Strengths and limitations Our study used large patient samples from two healthcare systems. VA is a nationwide system that cares for military veterans, most of whom are men and tend to have lower incomes and higher comorbidity burdens. 36 37 We focused on older veterans, whose prevalence of diabetes is about 25%12 and updated the cohort from our prior studies.16–18 We supplemented VA data with Medicare claims to enhance the completeness of diagnoses and outcomes. We also used comprehensive healthcare data from two separate KP health systems that contributed patients with more diverse characteristics. A1c TIR was calculated over a baseline period followed by the outcomes assessment to minimize risks of reverse causation. Our adjusted models controlled for many patient-level factors and clinical characteristics. We employed IV models to further limit confounding and used a falsification sample to test the validity of the IV. To fully address the causal question of A1c TIR on major outcomes would require a randomized trial, but observational studies using advanced statistical techniques may help define such estimates.Our study also has several limitations. First, our requirement for four or more A1c tests during baseline may lead to selection bias for those who have more interactions with the healthcare system. This may also result in surveillance bias, whereby the frequency of A1c measurements and A1c TIR calculations is influenced by the patient’s health status. Second, some factors that may affect A1c stability are not coded in electronic health records, such as diabetes duration or engagement with self-care, and cannot be included in the analyses. Third, few patients were receiving newer diabetes medications, such as sodium–glucose cotransporter-2 inhibitors or glucagon-like peptide-1 receptor agonists, which have been shown to reduce cardiovascular and renal outcomes. It is unclear if the risks of low A1c TIR are mitigated with these medications. Fourth, to estimate outcomes, we used Cox proportional hazards models, which assume that relative risk is constant over time. These models are generally robust even when such assumptions are violated.38 Nonetheless, we also assessed mortality using accelerated failure time models, which showed very similar results. Fifth, to maintain separation between the exposure and outcome periods, our study design did not track time-varying covariates during participant follow-up, such as medications, risk factors, or A1c levels, which may affect outcomes. Sixth, as an observational study, there are likely to be some unmeasured factors that cannot be controlled. From a comparison of the adjusted Cox and IV models in KP and VA patients, the influence that unmeasured factors have on the estimated effects may be somewhat different between the two populations. Finally, although we used quasi-experimental techniques to control for confounding, we cannot claim causality between A1c TIR, mortality, and diabetes complications or assert that prospectively increasing A1c TIR as a treatment strategy reduces these outcomes. Further research should examine these areas of uncertainty.Conclusions In conclusion, higher A1c TIR within patient-specific target ranges is associated with lower risks of mortality and cardiovascular outcomes among older adults with diabetes from two large health systems. These results confirm the role of A1c stability as a risk predictor for major adverse outcomes in older adults with diabetes.