This paper is in the following e-collection/theme issue:

Glucose Tracking and Self-Monitoring of Blood Glucose (150) Machine Learning (1383)

Published on in Vol 4, No 3 (2019): Jul-Sep

Preprints (earlier versions) of this paper are available at https://preprints.jmir.org/preprint/12905, first published .
A Reinforcement Learning–Based Method for Management of Type 1 Diabetes: Exploratory Study

A Reinforcement Learning–Based Method for Management of Type 1 Diabetes: Exploratory Study

A Reinforcement Learning–Based Method for Management of Type 1 Diabetes: Exploratory Study

Authors of this article:

Mahsa Oroojeni Mohammad Javad1 Author Orcid Image ;   Stephen Olusegun Agboola2, 3 Author Orcid Image ;   Kamal Jethwani2 Author Orcid Image ;   Abe Zeid4 Author Orcid Image ;   Sagar Kamarthi4 Author Orcid Image
Article Authors Cited by (16) Tweetations (3) Metrics

    Original Paper

    1Department of Information Technology and Analytics, Kogod School of Business, American University, Washington, DC, United States

    2Department of Dermatology, Harvard Medical School, Boston, MA, United States

    3Partners HealthCare, Boston, MA, United States

    4Mechanical and Industrial Engineering Department, College of Engineering, Northeastern University, Boston, MA, United States

    Corresponding Author:

    Mahsa Oroojeni Mohammad Javad, PhD

    Department of Information Technology and Analytics

    Kogod School of Business

    American University

    Kogod School of Business Building

    4400 Massachusetts Ave NW

    Washington, DC, 20016

    United States

    Phone: 1 202 885 2698

    Email: oroojeni@american.edu


    Background: Type 1 diabetes mellitus (T1DM) is characterized by chronic insulin deficiency and consequent hyperglycemia. Patients with T1DM require long-term exogenous insulin therapy to regulate blood glucose levels and prevent the long-term complications of the disease. Currently, there are no effective algorithms that consider the unique characteristics of T1DM patients to automatically recommend personalized insulin dosage levels.

    Objective: The objective of this study was to develop and validate a general reinforcement learning (RL) framework for the personalized treatment of T1DM using clinical data.

    Methods: This research presents a model-free data-driven RL algorithm, namely Q-learning, that recommends insulin doses to regulate the blood glucose level of a T1DM patient, considering his or her state defined by glycated hemoglobin (HbA1c) levels, body mass index, engagement in physical activity, and alcohol usage. In this approach, the RL agent identifies the different states of the patient by exploring the patient’s responses when he or she is subjected to varying insulin doses. On the basis of the result of a treatment action at time step t, the RL agent receives a numeric reward, positive or negative. The reward is calculated as a function of the difference between the actual blood glucose level achieved in response to the insulin dose and the targeted HbA1c level. The RL agent was trained on 10 years of clinical data of patients treated at the Mass General Hospital.

    Results: A total of 87 patients were included in the training set. The mean age of these patients was 53 years, 59% (51/87) were male, 86% (75/87) were white, and 47% (41/87) were married. The performance of the RL agent was evaluated on 60 test cases. RL agent–recommended insulin dosage interval includes the actual dose prescribed by the physician in 53 out of 60 cases (53/60, 88%).

    Conclusions: This exploratory study demonstrates that an RL algorithm can be used to recommend personalized insulin doses to achieve adequate glycemic control in patients with T1DM. However, further investigation in a larger sample of patients is needed to confirm these findings.

    JMIR Diabetes 2019;4(3):e12905

    doi:10.2196/12905

    Keywords

    type 1 diabetes mellitus; T1DM; diabetes treatment; machine learning; reinforcement learning; Q-learning; insulin dose prescription 


    Background

    According to the 2017 national diabetic statistics report, diabetes was the seventh leading cause of death in 2015 and a major cause of cardiovascular and renal diseases in the United States [1]. The Centers for Disease Control and Prevention reports that the number of Americans with diabetes is predicted to double or triple by 2050. In 2015, 30.3 million people in the United States (9.4% of the population) had diabetes. Of these, about 1.25 million were reported to have type 1 diabetes mellitus (T1DM) [2,3]. In T1DM, the beta cells responsible for producing insulin in the pancreas are deficient because of autoimmune destruction. T1DM patients depend on lifelong insulin therapy, delivered by injection or a pump, for glycemic control. Uncontrolled blood sugar can lead to serious short-term problems, such as hypoglycemia, hyperglycemia, or diabetic ketoacidosis [1,4-6], or chronic problems that can damage blood vessels supplying blood to important end organs, such as the heart, kidneys, eyes, and nerves [7,8]. Management of T1DM and its complications is achieved via pharmacotherapy, exercise, diet, and other lifestyle changes [9,10]. As individual patients have different physiological characteristics, they respond differently to treatments. Therefore, personalized treatment planning is likely to offer a more effective solution to managing glucose level and diabetes complications.

    Literature Review

    Some studies analyzed diabetes data and built models to predict blood glucose level [11-13]. Breault et al (2002) applied a classification and regression tree on data from 15,902 patients with diabetes to predict blood glucose level [14]. Yamaguchi et al (2006) used data collected over a period of 150 days from patients with T1DM to predict next-day-morning fasting blood glucose. They considered metabolic rate, food intake, and physical conditions as predictor variables and concluded that the physical conditions were highly correlated with fasting blood glucose [15]. Bellazzi et al (1998) used a combination of structural time series analysis and temporal abstraction for interpreting historic blood glucose level to extract and visualize the trends and daily cycles of blood glucose level [16]. Bellazzi and Abu-Hanna (2009) applied a temporal abstraction and subgroup discovery algorithm for predicting the blood glucose level of diabetes for 2 types of patients: those who self-monitor their blood glucose level at home and those who were admitted to an intensive care unit [17].

    Many studies have used computer-based systems, including open-loop and closed-loop control systems, to control the blood glucose levels of patients with diabetes. In the open-loop system, the patient or diabetologist is responsible for decision making regarding administration of each insulin injection [18]. On the other hand, the closed-loop system mimics the function of the pancreas to control blood glucose level [16-18]. A closed-loop system for T1DM includes either a model-free or a model-based method [19], which follows a cycle of steps: blood glucose measurement, insulin demand calculation, and insulin injection [18]. Many researchers attempted to use model-based control techniques to solve problems associated with diabetes [20,21]. Few studies applied a reinforcement learning (RL) algorithm for controlling blood glucose for type 1 diabetes.

    Only a few studies have applied model-based RL algorithm for controlling blood glucose levels for type 1 diabetes. Vrabie et al (2018) proposed using RL for obtaining optimal adaptive control algorithms for dynamical systems using the mathematical models [22]. Ngo et al (2018) used an RL-based algorithm for optimal control of blood glucose in patients with type 1 diabetes using simulations on a combination of the minimum model and part of the Hovorka model [23]. Ngo et al (2018) proposed an RL algorithm for automatically calculating the basal and bolus insulin doses for type 1 diabetes patients using simulation on a blood glucose model with Kalman filter [24].

    Currently, there are no effective algorithms to automatically control insulin delivery considering the blood glucose level feedback from the patient body. Only a few studies have attempted a data-driven approach to find a solution. Albisser et al (1974) applied a data-driven approach for developing artificial pancreas based on data from only 3 patients [25]. Javad et al (2015) proposed an RL approach for insulin dosage recommendation for patients with T1DM using an insulin pump based on the data from limited number of patients and states [26].

    In this study, we use a data-driven approach where an RL agent learns the model from patient data. The main purpose of this paper is to explore an RL-based approach to recommend personalized treatment plan for managing glucose level to prevent diabetes-related complications and improve quality of life in patients with T1DM.

    Overview of Reinforcement Learning

    RL discovers a policy to map a situation to an action to maximize a numeric reward, which takes into consideration not only the immediate rewards but also the possible subsequent rewards (delayed rewards) leading to an outcome such as a state where blood glucose is controlled. An RL agent determines which actions lead to the best reward through exploration of state space and exploitation of experience [27,28]. It has been applied successfully in different scientific fields such as robotics and control [29], manufacturing, and combinatorial search problems such as computer games [30,31]. In health care, using medical image and treatment regimen–related information from historical medical data, RL was used for cancer prediction, diagnosis, and prognosis [32,33].

    In RL, the learner or decision maker is called an agent (Q-learning in this application; it is described in the Methods section) that interacts with an environment (patient with T1DM in this application). Other 4 main subelements of RL include a policy (prescription medication level for a given patient condition in this application), a reward function (which estimates the reward, either positive or negative, depending on whether or not HbA1c level was controlled), a value function (Q-table in this application), and optionally, a model of the environment (not used in this application). In this application, let S be the set of all possible states of the environment (states of the T1DM patient) and A be the set of all possible actions (actions are the insulin levels prescribed to treat the T1DM patient). At each sequence of discrete time steps t=0,1,2,3,…, the RL agent receives a representation of the environment’s state st∈S. Considering available actions when environment is in state st, the agent takes an action at∈A, randomly at the early exploratory learning stage and more rationally exploiting the experience gained through data-driven learning in the advanced learning stage. The RL agent, depending on the consequence of its action at time t, receives a numerical reward rt and changes the environment to state st+1. Normally, the merit of an action is quantified by the total amount of reward that the RL agent can expect to accumulate in the long run, considering the states that are likely to be visited in the transition. Over a series of learning epochs, the RL agent learns an optimal control policy π*: S →A. At each time step time t, the optimal policy π*(st) maps state st to a right action at, that is, at= π*(st). Figure 1 shows the agent-environment (agent-patient) interaction in RL. The optimal control policy is shaped through exploration in the early stages of learning and through experience in the mature stage of learning.

    In this study, we apply a data-driven model-free RL method, known as Q-learning, that needs no previous knowledge of the environment to prescribe medication dose to treat T1DM patients considering their current HbA1c, body mass index (BMI), activity level, and alcohol usage.

    ‎
    Figure 1. The agent-environment interaction in reinforcement learning.
    View this figure

    This section describes Q-learning as applied to T1DM and its components including parameters that define state space and action space, reward function, training processes, training data, and evaluation function.

    Q-Learning

    Q-learning is useful for finding optimal strategies for an environment for which neither the transition function nor the probability distribution of state variables is known [34]. Q-learning works by estimating a set of Q-values, which serves as the role of a value function. In the Q-learning algorithm, Q-values are estimated for each state-action (st,at) combination. Once the final Q-values are estimated, the only thing that needs to be known is the state of the environment (T1DM patient) st to determine a right action at (insulin dose).

    At the beginning of the algorithm, Q-values are initiated to an arbitrary real number. Subsequently, at each iteration t, for each combination of state st∈S and action at∈A, a reward value is calculated by the RL agent. At the core of the algorithm is the iterative process of updating Q-values as a function of the immediate reward rt and Q-values of the next state-action pair Q(st+1, at+1). Figure 2 shows Q-value update function.

    In the above formulation, γ is a factor that regulates the influence of the future rewards relative to the current reward. If γ=0, the reward only depends on the reward received in the current state; as γ approaches 1, the reward is maximized over the long run taking future rewards into consideration [27,28]. Over several iterations of learning, Q-values for state-action pair, Q(st, at), converge to stable values and the RL agent is considered to have learned the optimal policy π*:S→A. At each time step time t, given state st, the right action at is determined from the formula presented in Figure 3.

    ‎
    Figure 2. Q-value update function.
    View this figure
    ‎
    Figure 3. Optimal policy function.
    View this figure

    Q-Learning Applied to Type 1 Diabetes Mellitus

    In this study, we study a Q-learning algorithm that prescribes medication level to a T1DM patient considering his or her state defined by HbA1c, BMI, activity level, and alcohol usage. The data for training Q-learning were obtained from electronic health records (EHRs) of patients admitted to the Mass General Hospital (MGH).

    Parameters That Define State Space

    On the basis of American Diabetes Association report, several factors such as diet, medication adherence, alcohol usage, physical activity, BMI, stress, age, smoking status, and side effects from other medications can change the blood glucose level of diabetes patients [1]. To identify the factors that are crucial for developing an effective machine learning model to personalize diabetes treatment planning, we calculated the correlation coefficient matrix of potential variables recorded in the EHR and observed that only BMI, activity level, and alcohol usage were strongly correlated with the blood glucose level measured in terms of HbA1c; other potential variables, such as age and smoking status, did not show significant correlation coefficients. Therefore, in this study, we defined a patient’s state by the 4 factors that influence the patient’s future HbA1c: current HbA1c, BMI, activity level, and alcohol usage.

    We denote the set of HbA1c states at epoch t by t={at|a=1,2,3}, the set of BMI levels by BMIt={BMIbt| b=1,…,17}, the set of activity levels by activity_levelt={activity_levelct| c=1,2}; and the set of alcohol usage levels by alcohol_usaget={alcohol_usagedt| d=1,2,3}. Table 1 presents the levels for HbA1c, BMI, activity level, and alcohol usage. The set of health states of a T1DM patient at epoch t is defined by st=(t, BMIt, activity_levelt, alcohol_usaget).

    Table 1. Definitions of levels for glycated hemoglobin, body mass index, activity level, and alcohol usage.
    VariableLevel 1Level 2Level 3Levels 4 to 16Level 17
    Glycated hemoglobin≤7: glucose level is well controlled(7,9]: glucose level is moderately controlled>9: glucose level is poorly controlledNAaNA
    Body mass index distribution[18.5,19)[19,20)[20,21)[21,22) to [33,34)(34,35]
    Activity levelActive—engages in physical activity ≥2 times per weekNonactive—engages in physical activity <2 times a weekNANANA
    Alcohol usageMild to no alcohol consumption—consumption of alcohol <2 times a weekModerate to high alcohol consumption—consumption of alcohol ≥2 times per weekHeavy consumption—consumption of alcohol few times a dayNANA

    aNot applicable.

    Parameters That Define Action Space

    Insulin is the mainstay of T1DM treatment and mostly administered through injections. The type of insulin that a T1DM patient needs depends on the severity of insulin depletion. There are different types of insulin used to treat T1DM. Normally, these insulin supplements are classified as short, rapid, intermediate, or long-acting. In this exploratory research, we focus only on the prescription of the most commonly prescribed long-acting insulin, that is, insulin glargine, which goes by the common brand name Lantus.

    Lantus is usually injected once per day at the same time each day. Once injected, Lantus works for about 24 hours. This is similar to the action of insulin normally produced by the pancreas to keep a patient’s blood sugar under control throughout the patient’s daily routine. Adding rapid-acting insulin to the long-acting background insulin prevents increasing a patient’s blood glucose right after eating a meal [7]. In the proposed Q-learning algorithm, actions represent the Lantus medication dosage levels recommended to the patients. Possible actions are coded based on 6 Lantus dosage ranges: a1t=[6,15), a2t=[15,20), a3t=[20,30), a4t=[30,40), a5t=[40,50), and a6t=[50,100]; these levels are referred to as Action 1, Action 2, …, Action 6, respectively. The set of possible actions at epoch t is denoted by at={ akt| k=1,2,…,6}, in other words, at={Action 1,Action 2,…,Action 6}. Actions are taken at a discrete decision epoch indexed by t= 1,2,..., T, where epoch t represents the time of the patient’s visit to physician’s office to get checkup and Lantus prescription. The patient’s visits (approximately every 3 months) to their physician over 10 years are treated as decision epochs.

    Reward Function

    In the proposed algorithm, the RL agent receives reward at each state comparable with the change in the state of HbA1c. At the beginning, the patient is in state s1 and takes treatment action a1; as a result, the agent receives reward r1 and the patient moves on to state s2; then the patient takes treatment a2, the agent receives reward r2, and the patient reaches state s3; and the procedure continues in this fashion. From a series of data-driven experiences, the RL agent learns the right action at (prescription of right Lantus dose) for a given patient state st. Figure 4 shows the reward function for the Q-learning algorithm.

    ‎
    Figure 4. Reward function.
    View this figure
    Training Processes

    In the training process, the Q-learning agent in this algorithm tries to learn the optimal treatment policy from the patient’s historical data in the EHR. At each iteration, the agent updates a table of Q-values for each combination of state and action. For example, each experience cycle (st, at, st+1, rt) updates the value of Q(st, at) according to the Equation 1. In this implementation, ε-greedy policy is applied for taking actions during the training process. Implementing ε-greedy policy helps the algorithm visit and explore different states by choosing random actions with small probability ε, instead of always taking experience-driven promising actions all the time. In this method, at each time step t, the algorithm selects a random action with a fixed probability, ε, based on the following formulation. Figure 5 shows the random action selection function, where 0≤ ut≤1 is a uniform random number drawn at each time step t [23,24].

    ‎
    Figure 5. Random action selection function.
    View this figure
    Training Data

    RL algorithm was trained and tested on the clinical data obtained from the MGH. The study was approved by the Partners Human Research Committee, the institutional review board that grants approval for such studies. In the dataset, most of the patients used Lantus compared with other types of insulin. So, this exploratory research focuses on only Lantus treatment planning for T1DM. Medical records of 87 T1DM patients enrolled at MGH from 2003 to 2013 were included in the training set. Only the patients who had complete data necessary for training the Q-learning agent were included in this analysis. Medical record data for each patient’s visits over a 10-year period were collected and processed for analyses. At each clinical encounter, HbA1c, BMI, activity level, alcohol usage status, and Lantus medication dose were recorded. Table 2 shows a sample of patient data collected from each visit. In addition, we validated the trained Q-learning agent performance on another dataset with 60 MGH patients for whom complete data were available.

    Evaluation Function

    Consider that (li, ui) is the Lantus dose interval recommend by the RL agent for test case I, and yi is the actual Lantus dose prescribed by the patient’s physician, and there are n number of cases in the validation set. The following equation was used for calculating the average error of RL agent predications. Figure 6 shows error function.

    Table 2. Tracking the patients’ visits.
    VisitHbA1ctBody_mass_indextActivity_leveltAlcohol_usagetLantus_doset
    18.121.41120
    29.1241122
    38221121
    ‎
    Figure 6. Error function.
    View this figure

    The average age of the study population was 53 years, 59% of the patients were male, 86% were white, and 47% were married. Table 3 shows demographics characteristics of patients included in the training data.

    Table 4 shows demographics characteristics of patients included in the testing data. Table 5 presents the results of Q-learning algorithm for 60 test cases. For the 60 test patients, on average, in 53 out of 60 cases (88%) the physician-prescribed Lantus dose was within the dose interval recommended by the Q-learning algorithm.

    Table 3. Summary of training data of patients (N=87).
    Patient characteristicsStatisticsa
    Age (years)

    		Mean (SD)52.9 (15.7)

    		Median54
    Race distribution, n (%)

    		White75 (86)

    		Hispanic or Latino7 (8)

    		Black3 (3)

    		Asian1 (1)

    		Not recorded1 (1)
    Marital status, n (%)

    		Married or partnered41 (47)

    		Single or widow33 (38)

    		Divorced or separated12 (14)

    		Widowed1 (1)
    Gender, n (%)

    		Male51 (59)

    		Female36 (41)

    aDue to rounding, the sum of the percentages shown is not 100.

    Table 4. Summary of test data of patients (N=60).
    Patient characteristicsStatistics
    Age (years) 
     Mean (SD)50.4 (15.8)
     Median52
    Race distribution, n (%a) 
     White53 (88)
     Hispanic or Latino5 (8)
     Not recorded2 (3)
    Marital status, n (%a) 
     Married or partnered32 (53)
     Single or widow22 (36)
     Divorced or separated6 (10)
    Gender, n (%) 
     Female34 (57)
     Male26 (43)

    aDue to rounding, the sum of the percentages shown is not 100.

    Table 5. Test results.
    Test numberHemoglobin A1c levelBody mass index levelActivity levelAlcohol usageActual Lantus Units dosage prescribedReinforcement learning agent–recommended Lantus Units dose intervalComparison of actual Lantus dose with reinforcement learning agent–recommended Lantus dose interval
    117116[6,15)match
    2121114[6,15)match
    3251112[6,15)match
    4271120[6,15)not match
    5241114[6,15)match
    6251110[6,15)match
    7261112[6,15)match
    8281120[20,30)match
    9241120[20,30)match
    102111125[20,30)match
    112101125[20,30)match
    12281113[20,30)not match
    13261110[6,15)match
    14221111[6,15)match
    15241112[6,15)match
    16251113[6,15)match
    17271122[6,15)not match
    1824118[6,15)match
    1924116[6,15)match
    2024119[6,15)match
    21251110[6,15)match
    22251114[6,15)match
    23281215[6,15)match
    242141220[20,30)match
    25281218[6,15)not match
    26251214[6,15)match
    27251214[6,15)match
    2825128[6,15)match
    29281215[6,15)match
    30251211[6,15)match
    31251210[6,15)match
    3224129[6,15)match
    33261320[20,30)match
    34251320[20,30)match
    35251320[20,30)match
    36161320[15,20)match
    373122146[30,40)not match
    38152115[15,20)match
    39172120[15,20)match
    40152115[15,20)match
    41282125[20,30)match
    422102130[20,30)match
    433132150[30,40)not match
    44282121[20,30)match
    45262120[20,30)match
    46282128[20,30)match
    47262120[20,30)match
    48272120[15,20)match
    493122150[50,100]match
    503172170[50,100]match
    513172180[50,100]match
    52282125[20,30)match
    53262130[20,30)match
    543122150[30,40)not match
    553112136[30,40)match
    563112138[30,40)match
    57282121[20,30)match
    58292123[20,30)match
    59292123[20,30)match
    60272120[20,30)match

    Principal Findings

    Alcohol usage, physical activity, BMI, stress, and HbA1c level are crucial for developing effective models to personalize diabetes treatment planning [1]. In this study, a Q-learning agent that predicts personalized insulin dosages was formulated, trained, and tested considering patients’ current HbA1c, BMI, activity level, alcohol usage to define the patient state at epoch t: st={t, BMIt, activity_levelt, alcohol_usaget}. In other words, a patient can be in any of the 306 possible states (number of HbA1c states*number of BMI states*number of activity level states*number of alcohol usage status states=3 × 17 × 2 × 3). Each of these combinations represents a state. For example, if the patient is in state st, the dosage recommendation at, appropriate to state st, is suggested by Q-learning agent for that patient. Q-learning agent–recommended Lantus dose interval includes the actual prescription dose in 88% of the cases.

    Limitations

    This research has several limitations. We did not include other important lifestyle information about patients’ diet, stress, and medication adherence. These are well-known factors that influence blood glucose levels but are infrequently documented in the medical records. We suggest considering these factors in future research for developing more effective blood glucose control. Another important limitation is the small training dataset. The main constraint to evaluating the model in a larger cohort of patients was the time it took to clean and extract these important but poorly documented factors. With adequate funding, we can apply more sophisticated natural language processing techniques to capture data from unstructured text or note from a larger sample of patients. Yet another factor is the limited generalizability of the study findings. Study data were from patients in a large academic medical center that has a diabetes center and access to other supportive lifestyle change programs that may not be available in community health centers. The fact that only 1 type of insulin (Lantus) was included broadly limits the application of this study. However, as a proof of concept, we demonstrated that this concept could potentially be used for other insulin regimen as well.

    Comparison With Previous Studies

    Although in recent years, we have seen increased interest in applying machine learning methodologies in the study of personalize diabetes treatment planning, this study is the first of its kind that aims at finding the best insulin dosage for the T1DM for several reasons. First, this study involved the use of crucial factors including alcohol usage, physical activity, BMI, and HbA1c level for finding the best insulin dosage for patients with type 1 diabetes. None of the earlier studies in the literature considered all of these important factors for developing effective models to personalize diabetes treatment planning. Second, 2 patients with the same BMI and HbA1c but different alcohol usage and activity level need different insulin dosages for managing their blood glucose level. Considering only BMI and HbA1c for insulin dosage recommendation may lead to suggesting the same dose of medication to patients with different insulin dosage needs. Finally, this study involved the use of a larger clinical dataset compared with other datasets used in other studies concerned with managing blood glucose level. Data gathered from clinical settings have an important and complementary role in the research outcomes. The suggested model-based approaches in the literature used mathematical models for simulating the function of pancreas. These model-based approaches did not consider patient’s alcohol usage and physical activity level for the insulin dosage recommendation.

    Yasini et al (2003) applied an agent-based simulation for managing blood glucose of patients with diabetes based only on blood glucose levels [19]. For each state of glucose level, their algorithm provided only 1 insulin dosage recommendation without considering the patient’s BMI, activity level, or alcohol usage. Our proposed algorithm provides more precise insulin dosage recommendation considering the patient’s current HbA1c, BMI, activity level, or alcohol usage. Vrabie et al (2018) and 2 studies by Ngo et al (2018) applied a model-based RL algorithm for controlling blood glucose for type 1 Diabetes [22-24]. We used a data-driven approach and considered the blood glucose level feedback from the patient body for training the Q-learning algorithm. In addition, our proposed Q-learning algorithm considers not only the blood glucose of the patient for the insulin dosage recommendation but also the patient’s current HbA1c, BMI, activity level, and alcohol usage. Javad et al (2015) applied data-driven approach on the limited number of patients and small dimension of problem with only 13 states for insulin dosage recommendation of type 1 diabetes, without testing the results [26]. Our proposed algorithm provides more precise insulin dosage recommendation based on the 306 possible patient states, and the results have been validated. RL algorithm was trained on the clinical data obtained from 87 T1DM patients enrolled at MGH from 2003 to 2013. Furthermore, the performance of the RL agent was evaluated on 60 test cases.

    Conclusions

    Effective decision making about correct insulin dose may delay or prevent diabetes complications, such as heart attack, kidney disease, blindness, and amputation [2]. Study findings suggest that physicians may be able to use a Q-learning agent that considers patients’ BMI, activity level, alcohol usage status, and current HbA1c level to recommend insulin doses. This machine learning model may help improve the timeliness of achieving an effective treatment dose rather than multiple dosage trials based on clinical acumen alone. In addition to improving treatment efficacy time, this has the potential to reduce patient stress (less clinic visits), reduce health care costs, and improve overall quality of life. Future research could extend this proof-of-concept Q-learning model to include other types of insulin and other types of diabetes medications and other state variables. The performance of the Q-learning model can be enhanced by considering finer categories and intervals for defining a patient state and action. It may also be worth exploring in patients with type 2 diabetes.

    Conflicts of Interest

    None declared.

    1. American Diabetes Association. Statistics About Diabetes: Overall Numbers, Diabetes and Prediabetes   URL: http://www.diabetes.org/diabetes-basics/statistics/ [accessed 2019-04-22]
    2. Centers for Disease Control and Prevention. CDC Newsroom   URL: https://www.cdc.gov/media/pressrel/2010/r101022.html [accessed 2019-04-22]
    3. Martinez-Millana A, Jarones E, Fernandez-Llatas C, Hartvigsen G, Traver V. App features for type 1 diabetes support and patient empowerment: systematic literature review and benchmark comparison. JMIR Mhealth Uhealth 2018 Nov 21;6(11):e12237 [FREE Full text] [CrossRef] [Medline]
    4. National Institute of Diabetes and Digestive and Kidney Diseases. Kidney Disease Statistics for the United States   URL: http:/​/www.​niddk.nih.gov/​healthinformation/​healthtopics/​kidneydisease/​kidney-disease-of-diabetes/​Pages/​facts.​aspx [accessed 2019-04-22]
    5. The United States Renal Data System. 2016 Annual Data Report   URL: https://www.usrds.org/adr.aspx [accessed 2019-04-22]
    6. Centers for Disease Control and Prevention. 2011. National Diabetes Fact Sheet   URL: https://www.cdc.gov/diabetes/pubs/pdf/ndfs_2011.pdf [accessed 2019-04-22]
    7. Mayo Clinic. Diabetes: Overview   URL: https://www.mayoclinic.org/diseases-conditions/diabetes/symptoms-causes/syc-20371444 [accessed 2019-04-22]
    8. Genuth S, Eastman R, Kahn R, Klein R, Lachin J, Lebovitz H, American Diabetes Association. Implications of the United kingdom prospective diabetes study. Diabetes Care 2003 Jan;26(Suppl 1):S28-S32. [CrossRef] [Medline]
    9. Woldaregay AZ, Årsand E, Botsis T, Albers D, Mamykina L, Hartvigsen G. Data-driven blood glucose pattern classification and anomalies detection: machine-learning applications in type 1 diabetes. J Med Internet Res 2019 May 1;21(5):e11030 [FREE Full text] [CrossRef] [Medline]
    10. Jeon E, Park H. Experiences of patients with a diabetes self-care app developed based on the information-motivation-behavioral skills model: before-and-after study. JMIR Diabetes 2019 Apr 18;4(2):e11590 [FREE Full text] [CrossRef] [Medline]
    11. Farmer TG, Edgar TF, Peppas NA. The future of open- and closed-loop insulin delivery systems. J Pharm Pharmacol 2008 Jan;60(1):1-13 [FREE Full text] [CrossRef] [Medline]
    12. Breault JL, Goodall CR, Fos PJ. Data mining a diabetic data warehouse. Artif Intell Med 2002;26(1-2):37-54. [CrossRef] [Medline]
    13. Wells BJ, Lenoir KM, Diaz-Garelli JF, Futrell W, Lockerman E, Pantalone KM, et al. Predicting current glycated hemoglobin values in adults: development of an algorithm from the electronic health record. JMIR Med Inform 2018 Oct 22;6(4):e10780 [FREE Full text] [CrossRef] [Medline]
    14. Yamaguchi M, Kaseda C, Yamazaki K, Kobayashi M. Prediction of blood glucose level of type 1 diabetics using response surface methodology and data mining. Med Biol Eng Comput 2006 Jun;44(6):451-457. [CrossRef] [Medline]
    15. Bellazzi R, Magni P, Larizza C, de Nicolao G, Riva A, Stefanelli M. Mining biomedical time series by combining structural analysis and temporal abstractions. Proc AMIA Symp 1998;1:160-164 [FREE Full text] [Medline]
    16. Bellazzi R, Abu-Hanna A. Data mining technologies for blood glucose and diabetes management. J Diabetes Sci Technol 2009 May 1;3(3):603-612 [FREE Full text] [CrossRef] [Medline]
    17. Carson ER, Deutsch T. A spectrum of approaches for controlling diabetes. IEEE Control Syst Mag 1992 Dec;12(6):25-31. [CrossRef]
    18. Takahashi D, Xiao Y, Hu F. A survey of insulin-dependent diabetes-part II: control methods. Int J Telemed Appl 2008;4:739385 [FREE Full text] [CrossRef] [Medline]
    19. Yasini S, Naghibi-Sistani MB, Karimpour A. Agent-based simulation for blood glucose control in diabetic patients. Int J Biomed Biol Eng 2009;3(9):260-267 [FREE Full text]
    20. Shimoda S, Nishida K, Sakakida M, Konno Y, Ichinose K, Uehara M, et al. Closed-loop subcutaneous insulin infusion algorithm with a short-acting insulin analog for long-term clinical application of a wearable artificial endocrine pancreas. Front Med Biol Eng 1997;8(3):197-211. [Medline]
    21. Pagurek B, Riordon JS, Mahmoud S. Adaptive control of the human glucose-regulatory system. Med Biol Eng 1972 Nov;10(6):752-761. [CrossRef] [Medline]
    22. Vrabie D, Vamvoudakis KG, Lewis FL. Optimal Adaptive Control and Differential Games by Reinforcement Learning Principles. London, UK: Institution of Engineering and Technology; 2012.
    23. Ngo PD, Wei S, Holubova A, Muzik J, Godtliebsen F. Reinforcement-Learning Optimal Control for Type-1 Diabetes. In: Proceedings of the International Conference on Biomedical & Health Informatics. 2018 Presented at: BHI'18; March 4-7, 2018; Las Vegas, NV, USA p. 333-336. [CrossRef]
    24. Ngo PD, Wei S, Holubová A, Muzik J, Godtliebsen F. Control of blood glucose for type-1 diabetes by using reinforcement learning with feedforward algorithm. Comput Math Methods Med 2018;2018:4091497 [FREE Full text] [CrossRef] [Medline]
    25. Albisser AM, Leibel BS, Ewart TG, Davidovac Z, Botz CK, Zingg W, et al. Clinical control of diabetes by the artificial pancreas. Diabetes 1974 May;23(5):397-404. [CrossRef] [Medline]
    26. Javad MO, Agboola S, Jethwani K, Zeid I, Kamarthi S. Reinforcement Learning Algorithm for Blood Glucose Control in Diabetic Patients. In: Proceedings of the International Mechanical Engineering Congress and Exposition. 2015 Presented at: ASME'15; November 13-19, 2015; Houston, Texas, USA. [CrossRef]
    27. Sutton RS, Barto AG. Reinforcement Learning: An Introduction. Cambridge, UK: MIT Press; 1998.
    28. Sutton RS, Barto AG. Reinforcement Learning: An Introduction. Cambridge, UK: MIT Press; 2018.
    29. Kober J, Bagnell JA, Peters J. Reinforcement learning in robotics: a survey. Int J Robot Res 2013 Aug 23;32(11):1238-1274. [CrossRef]
    30. Berny A. Selection and Reinforcement Learning for Combinatorial Optimization. In: Proceedings of the International Conference on Parallel Problem Solving from Nature. 2000 Presented at: PPSN'00; September 18-20, 2000; Paris, France p. 601-610. [CrossRef]
    31. van Eck NJ, van Wezel M. Application of reinforcement learning to the game of Othello. Comput Oper Res 2008 Jun;35(6):1999-2017. [CrossRef]
    32. Gottesman O, Johansson F, Komorowski M, Faisal A, Sontag D, Doshi-Velez F, et al. Guidelines for reinforcement learning in healthcare. Nat Med 2019 Jan;25(1):16-18. [CrossRef] [Medline]
    33. Esteva A, Robicquet A, Ramsundar B, Kuleshov V, DePristo M, Chou K, et al. A guide to deep learning in healthcare. Nat Med 2019 Jan;25(1):24-29. [CrossRef] [Medline]
    34. Mitchell TM. Machine Learning. Boston, Massachusetts: McGraw Hill Education; 1997.

    ‎
    BMI: body mass index
    EHR: electronic health record
    HbA 1c: glycated hemoglobin
    MGH: Mass General Hospital
    RL: reinforcement learning
    T1DM: type 1 diabetes mellitus

    Edited by G Eysenbach; submitted 24.11.18; peer-reviewed by R Kim, H Yang, L Zhou, J Obeid, J Forte, A Louren; comments to author 20.06.19; revised version received 24.06.19; accepted 19.07.19; published 28.08.19

    Copyright

    ©Mahsa Oroojeni Mohammad Javad, Stephen Olusegun Agboola, Kamal Jethwani, Abe Zeid, Sagar Kamarthi. Originally published in JMIR Diabetes (http://diabetes.jmir.org), 28.08.2019.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Diabetes, is properly cited. The complete bibliographic information, a link to the original publication on http://diabetes.jmir.org/, as well as this copyright and license information must be included.