Evaluation of the Effectiveness of Remote Foot Temperature Monitoring for Prevention of Amputation in a Large Integrated Health Care System

Monitoring plantar skin temperatures, which has previously been shown to reduce the risk of diabetic foot ulcer (DFU) recurrence (1–3), is recommended by several organizations (4–6). New technologies (e.g., “smart” thermometric mats) intended to be used in the patient’s home can detect “hot spots,” which when addressed early (e.g., through off loading, activity reduction, and clinical evaluation and treatment), can prevent ulcer development. A single site pre-post study in 77 people found that SmartMat use was associated with a 91% relative risk reduction (RRR) in moderate and severe DFU, 71% RRR in amputations, and 52% RRR in all-cause hospitalizations (7). To our knowledge, no randomized controlled trials or multisite observational studies have been conducted to evaluate the effectiveness of remote foot temperature monitoring (RTM). The Veterans Health Administration (VHA), which cares for >2 million patients with diabetes, is an ideal environment for evaluating effectiveness of RTM. This study aimed to evaluate the effectiveness of RTM in the VHA in preventing amputation. Secondary outcomes were all-cause hospitalization and death.

We conducted a matched retrospective longitudinal cohort study using national VHA electronic health records. This study was classified as quality improvement and, thus, did not undergo Institutional Review Board review or approval.

We used the VHA’s Corporate Data Warehouse to extract demographic information; outpatient, inpatient, and emergency department use; inpatient and outpatient diagnosis and procedure codes; and date of death, when applicable. Podimetrics, Inc. provided data on average monthly use of the RTM mats.

Patients with a history of ulceration, osteomyelitis, lower-extremity amputation (LEA), or Charcot foot were eligible for RTM based on VHA guidance in place at the time of the study (K.N., personal communication). See Fig. 1 and Supplementary File 1 for more details on inclusion/exclusion criteria.

We classified patients as being enrolled in RTM if their medical records indicated they had been ordered a SmartMat between January 2019 and February 2021.

Patients enrolled in RTM are mailed a SmartMat and trained to use it in their home. Patients are instructed to place bare feet on the SmartMat for 20 s daily to take a temperature scan. The data are transmitted to the company via cellular technology and analyzed to detect hot spots based on intra- (for those with a single monitored foot) or interfoot temperature asymmetries (8,9). Prior research determined that temperature asymmetries of >2.2°C over two consecutive scans detected 91–97% of DFUs an average of 37–41 days before appearance of an ulcer (8,9). Per VHA protocols, when the 2.2°C threshold is exceeded for ≥2 days, a SmartMat care management team member or a VHA health care provider follows up with the patient. The RTM device has a warranty and lifecycle of 1 year.

All patients, including those enrolled in RTM, received usual care (10).

We matched, without replacement, up to three eligible patients not enrolled in RTM to every RTM patient based on factors noted in Table 1.

For the main analyses, outcome ascertainment was between January 2019 and August 2021. The primary outcome was LEA at any level (see Supplementary Table 1 for codes). All-cause hospitalization was based on the presence of inpatient records and length of stay >1 day. Death was determined from having a recorded date of death.

Analyses adjusted for factors (measured in the 2 years prior to baseline) known to be associated with LEA: race/ethnicity, hemoglobin A_1c, osteomyelitis, Charcot foot, kidney disease, Gagne comorbidity index, drive time to specialty care, and emergency department/urgent care visits.

Data were incomplete for <0.5% of the sample. The main analysis excluded those with missing data.

We used conditional Cox regression accounting for matched data, stratifying on matched pairs (allowing baseline hazards to vary among matched pairs), to estimate unadjusted hazard ratios (HRs) and adjusted HRs (aHRs) and corresponding 95% CIs. For nondeath outcomes (LEA and all-cause hospitalization), cause-specific hazard was estimated, treating death as a competing risk, with the outcome of interest censored when death occurred first. We examined plots of scaled Schoenfeld residuals for substantial violations of the proportional hazards assumption as well as survival curves for each outcome.

Numerous sensitivity analyses to assess the robustness of findings were conducted and are described in Supplementary File 1.

We included 924 RTM enrollees (Fig. 1) who were matched with up to 3 patients not enrolled in RTM (n = 2,757 total). The RTM and comparison groups were similar on the matching factors. A greater percentage of patients enrolled in RTM had a history of osteomyelitis (41% vs. 33%), whereas 92% in both groups had a history of ulceration. At baseline, nearly 70% had no history of LEA, 24% had a prior toe or partial foot amputation, and 7% had a prior amputation at the ankle or more proximal (major LEA) (Table 1). A greater percentage of the comparison group had Care Assessment of Need scores (11) in the 99th percentile (4.9% vs. 2.8%), indicating that comparison group patients had a higher predicted risk of hospitalization and death.

The incidence rate of LEA was nearly the same in the RTM and comparison groups (4.8 vs. 4.9 per 100 person-years, respectively) (Table 2). Among the LEAs, 86% were minor (toe/partial foot) in the RTM group and 82% were minor in the comparison group. After covariate adjustment, RTM enrollment was not associated with LEA incidence (HR 0.92, 95% CI 0.62–1.37). The incidence rate of hospitalization was also similar between groups (34.1 and 35.6 per 100 person-years; aHR 0.97, 95% CI 0.82–1.14). Death was inversely associated (reduced risk) with RTM enrollment (aHR 0.63, 95% CI 0.49–0.82).

Sensitivity analyses were consistent with primary analyses (Supplementary Table 2 and Supplementary File 1).

Neither LEA nor all-cause hospitalization was associated with RTM enrollment, although RTM enrollment was associated with a lower risk of death. We hypothesized that reductions in hospitalizations and death would be mediated through reductions in LEA; because we only observed a reduction in death, it is more challenging to interpret our results. Patients enrolled in RTM may have had more contact with clinicians because of their enrollment, and these contacts may have provided opportunities to voice concerns about health issues and treat conditions (not limited to foot problems), reducing the risk of death, and possibly increasing treatments (including hospitalizations) in this group. Alternatively, providers may have selectively enrolled patients in RTM who had a lower risk of death based on characteristics that were not visible to us through the electronic health record data used for the study.

Because our finding for LEA was not as hypothesized, we conducted numerous post hoc sensitivity analyses to investigate possible sources of bias and evaluate whether some groups may receive more benefit from RTM than others. None of the sensitivity analyses for LEA indicated a statistically significant difference in risk. Surprisingly, those who used the SmartMat regularly (compliant) did not have a lower hazard for LEA relative to the comparison group, although HRs for death were lower among those who were compliant. Unmeasured factors associated both with SmartMat use and risk of death may explain the associations observed. Alternatively, poor health may cause patients to reduce SmartMat use.

There are several possible explanations for the weaker results in this study relative to prior studies (1–3,7). Patients may not have used the SmartMat as directed. Approximately 54% of the time, patients used the SmartMat at least twice per week, but there was no evidence that more consistent use was associated with a greater reduction in LEA risk. Patients may not have been alerted when there was a hot spot or not complied with instructions when alerted. The VHA guidelines in place at the time of this study did not specify what should happen after a temperature asymmetry was identified. Consequently, there is thought to be wide variation in responses, which was not captured but would be important to explore in the future. Selection bias is possible, but we matched on numerous factors known to be associated with our outcomes and adjusted for other factors.

Limitations for this study include the retrospective design, inability to assess ulceration (because codes cannot distinguish incident from prevalent ulcers), misclassification due to errors and/or omissions in electronic health record data, variation in usual care across facilities and regions, resulting in unmeasured confounding, and generalizability to nonveteran populations. Additionally, we were unable to exclude individuals with an active ulcer at baseline. Since amputation risk is higher in those with active ulcers (12), disproportionate inclusion of individuals with active ulcers in either group could lead to an elevated risk of amputation in that group, resulting in biased estimates of association.

In conclusion, our study does not provide support that RTM reduces risk of LEA or all-cause hospitalization in individuals with a history of DFU. Prospective studies and randomized controlled trials are needed to overcome important limitations and identify implementation gaps that can be filled to improve effectiveness.

Acknowledgments. The authors are grateful for the input of Alexander Peterson, Seattle Epidemiologic Research and Information Center, Department of Veterans Affairs Puget Sound Health Care System, Seattle, WA.

Funding. This work was supported in part by the Department of Veterans Affairs Health Services Research and Development and the Veterans Affairs Office of Health Equity (CIN 13-402).

The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the U.S. government.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.J.L. contributed to conceptualization, supervision, project administration, funding acquisition, writing the original draft, and reviewing and editing the manuscript. A.K.T. contributed to conceptualization, formal analysis, software, methodology, visualization, and reviewing and editing the manuscript. A.K. contributed to conceptualization, analysis, methodology, and reviewing and editing the manuscript. K.C.G.C. contributed to conceptualization, analysis, methodology, and reviewing and editing the manuscript. K.T.J. contributed to formal analysis, software, methodology, data curation, and reviewing and editing the manuscript. S.S. contributed to conceptualization and reviewing and editing the manuscript. K.N. contributed to conceptualization, analysis, and reviewing and editing the manuscript. J.R. contributed to conceptualization and reviewing and editing the manuscript. S.M. contributed to conceptualization, analysis, and reviewing and editing the manuscript. E.M. contributed to conceptualization, analysis, and reviewing and editing the manuscript. A.J.L. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the VA’s Health Services Research and Development meeting, Baltimore, MD, 8–10 February 2023, and at the Ninth International Symposium of the Diabetic Foot, The Hague, the Netherlands, 10–13 May 2023.