Home

Row

Welcome

My name is Karandeep Singh. I lead the Machine Learning for Learning Health Systems (ML4LHS) Lab. Welcome to our home page.

I am an Assistant Professor in the Departments of Learning Health Sciences and Internal Medicine at the University of Michigan Medical School and Assistant Professor of Information in the University of Michigan School of Information. I am a physician, researcher, and educator interested in applying machine learning methods to understand disease epidemiology and predict health outcomes. I teach a course on exploratory data analysis in health to graduate students. I completed my internal medicine residency at UCLA Medical Center, where I served as chief resident, and a nephrology fellowship in the combined Brigham and Women’s Hospital/Massachusetts General Hospital program in Boston, MA. I completed my medical education at the University of Michigan Medical School and hold a master’s degree in medical sciences in Biomedical Informatics from Harvard Medical School. I am board certified in internal medicine, nephrology, and clinical informatics.

Connect: Faculty Profile | LinkedIn | MCommunity

Welcome


My name is Karandeep Singh. I lead the Machine Learning for Learning Health Systems (ML4LHS) Lab. Welcome to our home page.

I am an Assistant Professor in the Departments of Learning Health Sciences and Internal Medicine at the University of Michigan Medical School and Assistant Professor of Information in the University of Michigan School of Information. I am a physician, researcher, and educator interested in applying machine learning methods to understand disease epidemiology and predict health outcomes. I teach a course on exploratory data analysis in health to graduate students. I completed my internal medicine residency at UCLA Medical Center, where I served as chief resident, and a nephrology fellowship in the combined Brigham and Women’s Hospital/Massachusetts General Hospital program in Boston, MA. I completed my medical education at the University of Michigan Medical School and hold a master’s degree in medical sciences in Biomedical Informatics from Harvard Medical School. I am board certified in internal medicine, nephrology, and clinical informatics.

Connect: Faculty Profile | LinkedIn | MCommunity

Row

What is ML4LHS?

The Machine Learning for Learning Health Systems (ML4LHS) Lab applies machine learning methods to understand disease epidemiology and predict health outcomes.

We have ongoing projects in several clinical domains, including:

  • Chronic kidney disease
  • Prostate cancer
  • Benign urology
  • Emergency care
  • Medication adherence and effectiveness

Row

Current Lab Members

Elliott Brannon, MPH

HILS PhD student

Connect: Student Profile | LinkedIn | MCommunity

Adharsh Murali

MHI student

Connect: LinkedIn | MCommunity

Etiowo Usoro

MHI student

Connect: Student Profile | LinkedIn | MCommunity

Tianshi Wang

MHI student

Connect: LinkedIn | MCommunity

Row

Lab Alumni

Shreyas Ramani, MHI

Healthcare Data Analyst, Caravan Health

Connect: LinkedIn

Pritika Dasgupta, MPH, MHI

Biomedical Informatics PhD Student, University of Pittsburgh School of Medicine

Connect: Student Profile | LinkedIn

Sajjad Seyedsalehi, MSc, MSc

Michigan State University

Connect: LinkedIn

Publications

Row

Publications

Click to view more details.
JM Franklin, C Gopalakrishnan, AA Krumme, K Singh, JR Rogers, …. The relative benefits of claims and electronic health record data for predicting medication adherence trajectory. American heart journal. 197, 153-162, 2018.
K Singh, H Wessells, JQ Clemens, RL Dunn, SK Holt, J Hotaling, …. MP04-01 LONGITUDINAL SUBPHENOTYPES OF URINARY SYMPTOMS IN TYPE I DIABETES. The Journal of Urology. 199 (4), e31-e32, 2018.
K Singh, KR Ghani, S Seyedsalehi, MH Solomon, G Auffenberg, …. MP16-14 THE UNPREDICTABILITY OF SOCIAL CONTINENCE AFTER RADICAL PROSTATECTOMY. The Journal of Urology. 199 (4), e202, 2018.
K Singh, AV Sarma, RL Dunn, SK Holt, J Hotaling, R Pop-Busui, …. MP74-17 TRAJECTORIES OF ERECTILE DYSFUNCTION SUBPHENOTYPES IN MEN WITH LONGSTANDING TYPE I DIABETES. The Journal of Urology. 199 (4), e1006, 2018.
DE Leaf, ED Siew, MF Eisenga, K Singh, FR Mc Causland, A Srivastava, …. Fibroblast Growth Factor 23 Associates with Death in Critically Ill Patients. Clinical Journal of the American Society of Nephrology. 13 (4), 531-541, 2018.
K Singh, AB Landman. Mobile Health. Key Advances in Clinical Informatics. 183-196, 2018.
K Singh, SS Waikar, L Samal. Evaluating the feasibility of the KDIGO CKD referral recommendations. BMC nephrology. 18 (1), 223, 2017.
GB Auffenberg, S Merdan, DC Miller, K Singh, BR Stockton, KR Ghani, …. Evaluation of prostate cancer risk calculators for shared decision making across diverse urology practices in Michigan. Urology. 104, 137-142, 2017.
JA Rodriguez, K Singh. The Spanish Availability and Readability of Diabetes Apps. Journal of diabetes science and technology. 1932296817749610, 2017.
TA Mavrakanas, A Khattak, K Singh, DM Charytan. Epidemiology and Natural History of the Cardiorenal Syndromes in a Cohort with Echocardiography. Clinical Journal of the American Society of Nephrology. 12 (10), 1624-1633, 2017.
K Singh, A Landman, DW Bates. For Reprints, Links &. Health Affairs. 36, 6, 2017.
K Singh, A Landman, DW Bates. Mobile Health Apps: The Authors Reply. Health Affairs. 36 (6), 1144-1144, 2017.
G Auffenberg, S Ramani, K Ghani, B Denton, C Rogers, B Stockton, …. PNFBA-06 ASKMUSIC©: LEVERAGING A CLINICAL REGISTRY TO INFORM PATIENTS. The Journal of Urology. 197 (4), e911, 2017.
K Singh, A Landman, DW Bates. Mobile Health Applications: The Authors Reply. Health Affairs. 36 (2), 384-384, 2017.
K Singh, K Drouin, LP Newmark, R Rozenblum, J Lee, A Landman, …. Developing a framework for evaluating the patient engagement, quality, and safety of mobile health applications. Issue Brief (Commonw Fund). 5 (1), 11, 2016.
U Sarkar, GI Gourley, CR Lyles, L Tieu, C Clarity, L Newmark, K Singh, …. Usability of commercially available mobile applications for diverse patients. Journal of general internal medicine. 31 (12), 1417-1426, 2016.
K Singh, K Drouin, LP Newmark, JH Lee, A Faxvaag, R Rozenblum, …. Many mobile health apps target high-need, high-cost populations, but gaps remain. Health Affairs. 35 (12), 2310-2318, 2016.
K Singh, K Drouin, LP Newmark, M Filkins, E Silvers, PA Bain, DM Zulman, …. Patient-facing mobile apps to treat high-need, high-cost populations: a scoping review. JMIR mHealth and uHealth. 4 (4), 2016.
MD Iversen, S Kiami, K Singh, I Masiello, J von Heideken. Prospective, randomised controlled trial to evaluate the effect of smart glasses on vestibular examination skills. BMJ Innovations, bmjinnov-. -2015-000094, 2016.
K Singh, A Wright. Clinical decision support. Clinical Informatics Study Guide. 111-133, 2016.
K Singh, RA Betensky, A Wright, GC Curhan, DW Bates, SS Waikar. A Concept–Wide Association Study of Clinical Notes to Discover New Predictors of Kidney Failure. Clinical Journal of the American Society of Nephrology. 11 (12), 2150-2158, 2016.
JM Franklin, AA Krumme, C Gopalakrishnan, K Singh, JR Rogers, …. Predicting Refill Adherence from Alternative Data Sources: Claims, Structured Electronic Health Record Data, and Physician Notes. PHARMACOEPIDEMIOLOGY AND DRUG SAFETY. 25, 190-191, 2016.
U Sarkar, GI Gourley, C Lyles, L Tieu, C Clarity, L Newmark, K Singh, …. MOBILE APPS FOR VULNERABLE POPULATIONS STUDY. JOURNAL OF GENERAL INTERNAL MEDICINE. 31, S303-S303, 2016.
G Auffenberg, S Merdan, D Miller, K Singh, B Stockton, B Denton, K Ghani. MP39-02 DEVELOPMENT AND VALIDATION OF A RISK PREDICTION MODEL FOR PROSTATE CANCER DIAGNOSIS WITHIN A STATEWIDE QUALITY IMPROVE…. The Journal of Urology. 195 (4), e541-e542, 2016.
GO Klein, K Singh, J von Heideken. Smart Glasses–A New Tool in Medicine.. Studies in health technology and informatics. 216, 901-901, 2015.
ML Mendu, LI Schneider, AA Aizer, K Singh, DE Leaf, TH Lee, SS Waikar. Implementation of a CKD checklist for primary care providers. Clinical Journal of the American Society of Nephrology, CJN.. 01660214, 2014.
K Singh, G Klein, J von Heideken. Google Glass has potential for rheumatology and orthopedic surgery. The Rheumatologist. 1-5, 2014.
BH Hahn, MA Mcmahon, A Wilkinson, WD Wallace, DI Daikh, …. American College of Rheumatology guidelines for screening, treatment, and management of lupus nephritis. Arthritis care & research. 64 (6), 797-808, 2012.

Learn More About Datasets

Row

Learn More About Datasets

The following datasets are currently available for download. Please click on the following links to learn more:

Mobile Health Apps Evaluation

Mobile Health Apps Evaluation

Row

Mobile Health Apps Evaluation Dataset

The primary dataset can be downloaded by clicking the link for data in the following format:

CSV

This is the publication associated with this dataset: Singh K, Drouin K, Newmark LP, et al. Many Mobile Health Apps Target High-Need, High-Cost Populations, But Gaps Remain. Health Aff. 2016;35(12):2310-2318. doi:10.1377/hlthaff.2016.0578.

What do the variables mean?

Note: Variable names containing spaces or punctuation had these removed to minimize compatability issues when importing for statistical analysis.

General

App ID: This is a unique identifier we assigned to each app. AppIDs ending in b were originally discovered as being present on both the Android and iOS app stores. Some of these may have been ultimately reviewed on only one of the 2 operating systems (OSes) if the app was pulled from one of the two app stores or if the app was non-functional on one of the 2 OSes.

Permission To Name: We asked app developers for permission to name their apps in our study. For more details, see the methods in the original manuscript.

y - the developer gave us permission to name their app

n - the developer did not give us permission to name their app

App Name: This is the name of the app as entered by the reviewer. Though we did not report the developer in the dataset, the names of the developers for each of the named apps are available in the appendix accompanying the manuscript.

Reviewer: Is the reviewer a clinician or non-clinician?

Options are: clinician_reviewer or nonclinician_reviewer.

Platform: This is the type of device on which the app was reviewed. Options are: iPhone, iPad or iPad Mini, Android phone, and Android tablet.

OS: This is the operating system on which the app was reviewed. This was derived automatically from the platform. Options are: iOS, Android, and Both.

Which populations is the app targeting?

Instructions to rater: Answer this question based on the app’s description in the app store and the app’s website. If the app is targeted at caregivers, try to determine the vulnerable populations who those individuals care for. Check ALL that apply. Some apps are very broad (like patient health records) and may be targeted at all conditions. For these types of apps, check only “All of the above.”

Options are: Asthma or COPD, Arthritis, Cancer, Chronic kidney disease, Cirrhosis or liver disease, Coronary artery disease or congestive heart failure, Dementia or mild cognitive impairment, Depression or bipolar disorder, Diabetes mellitus, Elderly, Hypertension, Obesity, Pain, Smoking, alcohol, or drug abuse, Stroke, All of the above, None of the above. The reason that None of the above was included as a response was to ensure that the question was not accidentally missed – at least one selection was required.

How would you best categorize the app developer?

Options are: Individual, For-profit company, Medical professional society, Other non-profit organization, and Government agency.

What kinds of support does the developer offer?

Instructions to rater: You may need to check the app store page and app website to determine the answer.

Seelct all that apply. Options are: E-mail, Phone, Live chat, Text messaging/SMS, None of the above. The reason that None of the above was included as a response was to ensure that the question was not accidentally missed – at least one selection was required.

Was an appropriate clinical expert involved in app development or qualitycontrol?

Instructions to rater: Answer this ONLY based on the app, app store page, and website. If you cannot determine this, choose No clinical expert involved.

Options are: Clinical expert involved and No clinical expert involved

Were patients involved in app development or quality control?

Instructions to rater: Answer this ONLY based on the app, app store page, and website. If you cannot determine this, choose No patients involved.

Options are: Patients involved and No patients involved

Does the app make reference to disease-specific guidelines?

Instructions to rater: Guidelines, broadly defined, may refer to professional society guidelines, review articles, or primary literature. If you cannot determine, please answer No.

Options are: Yes and No.

Does the app claim to be compliant with the FDA guidelines surrounding regulation of mobile medical apps?

Instructions to rater: Not all apps are required to be registered as medical devices but should acknowledge their standing with the FDA guidelines. If you cannot determine this, answer No.

Options are: Yes and No.

App Functionality

Does the app require a login?

Instructions to rater: Select yes if the app has no or minimal functionality without users creating an account. Otherwise, select no.

Does the app have in-app advertising?

Options are: Yes and No.

If you answered Yes to the prior question, is it possible to upgrade to an ad-free version?

Options are: Yes, No, and Not applicable (answered “No” to prior question)

Is the app tethered to a healthcare system?

Instructions to rater: Does the app require patients to be receiving care from a participating clinic or hospital in order to use the app?

Can the app sync directly with a peripheral device?

Instructions to rater: Examples include wearables (e.g. Fitbit) and other devices (e.g. thermometers, glucometers, and scales). If transferring data between a peripheral device and the app is a highly manual process (exporting a CSV file and reimporting a CSV file), then this does not count (answer No).

If the app is an iOS app, is it compatible with HealthKit?

Instructions to rater: This info should be on the app page description on the app store.

Options are: Yes, No, Not applicable (Android app)

If the app is an Android app, is it compatible with Google Fit?

Options are: Yes, No, Not applicable (iOS app)

Privacy and Security

Does the app have a privacy policy?

Instructions to rater: You may need to look on app page on the app store and on the app’s website. If you cannot find one, answer No.

Does the app developer provide user information to others individually or in aggregate?

Instructions to rater: This should be addressed in the privacy policy. Again, this may require looking at the app’s website.

Options are: Yes, individual information, Yes, in aggregate only, No, Not stated.

Does the app claim to meet the standard of HIPAA compliance?

Instructions to rater: You may need to look on app page on the app store and on app website. If you cannot determine this, answer No.

Patient Engagement

In what ways does the app engage patients?

Select all that apply. Options are: Provides educational information, Instructs patients (app provides guidance or advice based on information you provide it), Records information, Displays patient’s health information, Reminds or alerts patients, Enables communication of information with clinician, Enables communication of information with family (e.g. caregiver), None of the above. The reason that None of the above was included as a response was to ensure that the question was not accidentally missed – at least one selection was required.

Does the app reward the user for engaging with the app or achieving health goals?

Instructions to rater: Rewards could be in the form of points, games, rankings, or social media shares (e.g., app posting achievements to Facebook).

Options are: Yes and No.

Does the app engage users through social media? If so, how?

Options are: Public (e.g. Twitter, Facebook, Pinterest, etc), Private (e.g. app connects you to other patients through its own social media platform), Through a web-based forum, None of the above. The reason that None of the above was included as a response was to ensure that the question was not accidentally missed – at least one selection was required.

Does the app appropriately warn users, caregivers, or clinicians when dangerous information is entered?

Instructions to rater: As an example, does the app respond appropriately to very low or high blood sugars or very depressed/suicidal mood? The warning should reflect a sense of urgency that matches the degree of danger.

Options are: Yes and No.

How does the app handle communication or sharing of information with CAREGIVERS?

Instructions to rater: If the only way for a caregiver to access patient data is to login using the patient’s username/password, this does not count as data sharing.

Options are: Caregiver has access to information using separate login, Patient can e-mail information to caregiver using app, Patient can text information to caregiver using app, None of the above. The reason that None of the above was included as a response was to ensure that the question was not accidentally missed – at least one selection was required.

How does the app handle communication or sharing of information with CLINICIANS?

Instructions to rater: If the only way for a clinician to access patient data is to login using the patient’s username/password, this does not count as data sharing.

Options are: Clinician has access to information using separate login, Patient can e-mail information to clinician using app, Patient can text information to clinician using app, Patient can enter information directly into electronic health record or patient portal using app, None of the above. The reason that None of the above was included as a response was to ensure that the question was not accidentally missed – at least one selection was required.

System Usability Scale

The System Usability Scale (SUS) is a simple, ten-item scale giving a global view of subjective assessments of usability. In adapting this scale for evaluating app usability, we replaced “system” with “app.” For each question, option ranges from 1 (Strongly disagree) to 5 (Strongly agree).

Note: This question was completed only by non-clinician reviewers.

  1. I think that I would like to use this app frequently
  2. I found the app unnecessarily complex
  3. I thought the app was easy to use
  4. I think that I would need the support of a technical person to be able to use this app
  5. I found the various functions in this app were well integrated
  6. I thought there was too much in consistency in this app
  7. I would imagine that most people would learn to use this app very quickly
  8. I found the app very cumbersome to use
  9. I felt very confident using the app
  10. I needed to learn a lot of things before I could get going with this app

Costs: As an example, if the assigned app is free but a “Pro” version is available for $4.99 with the option of a subscription to premium content for $10/month, enter 0 in the first question below, 4.99 in the second question, and 120 in the last question.

Net Promoter Score

How likely is it that you would recommend this app to a friend or colleague?

Note: This question was completed only by clinician reviewers.

Options range from 0 (not likely at all) to 10 (extremely likely).

Consumer Ratings

iOS Avg Rating: This is the mean rating of all versions of an app as abstracted from the app store.

iOS Num Ratings: The is the total number of ratings for all versions of an app as abstracted from the app store.

Android Avg Rating: This is the mean rating of all versions of an app as abstracted from the app store.

Android Num Ratings: The is the total number of ratings for all versions of an app as abstracted from the app store.

Android Num Downloads: The is the total number of downloads for all versions of an app as abstracted from the app store.

Costs

iOS ONLY: Price to download the app (in dollars)

Instructions to rater: If the app is free, enter 0. This is price to download the app only (not including in-app purchases/subscriptions). Enter a NUMBER only (no dollar sign).

iOS ONLY: Price to purchase ALL in-app purchases (in dollars)

Instructions to rater: If no in-app purchases, enter 0. If there are multiple in-app purchases, please add up their cost to get this number. Enter a NUMBER only (no dollar sign).

iOS ONLY: Annual price to purchase a subscription (in dollars)

Instructions to rater: If no subscriptions are offered, enter 0. If monthly subscription cost is available, multiply by 12 to obtain annual price. Enter a NUMBER only (no dollar sign).

ANDROID ONLY: Price to download the app (in dollars)

Instructions to rater: If the app is free, enter 0. This is price to download the app only (not including in-app purchases/subscriptions). Enter a NUMBER only (no dollar sign).

ANDROID ONLY: Price to purchase ALL in-app purchases (in dollars)

Instructions to rater: If no in-app purchases, enter 0. If there are multiple in-app purchases, please add up their cost to get this number. Enter a NUMBER only (no dollar sign).

ANDROID ONLY: Annual price to purchase a subscription (in dollars)

Instructions to rater: If no subscriptions are offered, enter 0. If monthly subscription cost is available, multiply by 12 to obtain annual price. Enter a NUMBER only (no dollar sign)