An Owner-Independent Investigation of Diabetes Alert Dog Performance

Abstract

Objective: To quantify Diabetes Alert Dog (DAD) performance by using owner-independent measures. Research Design and Methods: Eight owners of accredited DADs used a FreeStyle Libre Flash Glucose Monitoring System (FGMS). Concurrent Closed Circuit Television (CCTV) footage was collected for between 5 and 14 days in each owner's home or workplace. The footage was blind-coded for dogs' alerting behaviors. The sensitivity, False Positive Rate and Positive Predictive Values (PPV) of dogs' alerts to out-of-range (OOR) episodes were calculated. Ratings for 11 attributes describing participant's lifestyle and compliance (taken from each dog's instructor) and the percentage of DAD alerts responded to by the owner as per training protocol (taken from CCTV footage) were assessed for association with dog performance. Results: Dogs alerted more often when their owners' glucose levels were outside vs. inside target range (hypoglycaemic 2.80-fold, p = 0.001; hyperglycaemic 2.29-fold, p = 0.005). Sensitivity to hypoglycaemic episodes ranged from 33.3 to 91.7%, the mean was 55.9%. Mean PPV for OOR episodes was 69.7%. Sensitivity and PPV were associated with aspects of the dog and owner's behavior, and the owner's adherence to training protocol. Conclusions: Owner-independent methods support that some dogs alert to hypo- and hyperglycaemic events accurately, but performance varies between dogs. We find that DAD performance is affected by traits and behaviors of both the dog and owner. Combined with existing research showing the perceived psychosocial value and reduced critical health care needs of DAD users, this study supports the value of a DAD as part of a diabetes care plan. It also highlights the importance of ongoing training and continued monitoring to ensure optimal performance.

Keywords: alert; behavior; canine; diabetes; hyperglycaemia; hypoglycaemia.

Figure 1

(A) Sensitivity for hypoglycaemic (left) and (B) hyperglycaemic episodes (middle) defined by each participant's own target range. (C) PPV for OOR episodes defined by each participant's target range (right). N = number of episodes, r = episodes with at least one alert during the episode or in the 15 min preceding it.

Figure 2

Individual dog's sensitivity to highs (hyperglycaemia) and lows (hypoglycaemia), and, when considering only those alerts once the owner is OOR, and when also including alerts 15 min prior to the first OOR recording. When two circles (one closed and one open) have the same dog number, this indicates a change in sensitivity when alerts in the processing 15 min are considered correct. Lines indicate the population mean (when including the preceding 15 min).

Figure 3

Clarke Error grids for each participant, comparing FreeStyle Libre FGMS with blood test results (% is percentage of data in zones A or B. Red crosses are data points outside these zones).

Figure 4

Odds ratios (95% confidence intervals) for potential predictor of sensitivity (to OOR episodes) and positive predictive value. Ratios are for a unit increase in score except for months since accreditation (per year), and % of alerts followed by blood test and percentage of alerts ignored (per 10 percentage point increase).