App Beats Paper

Reducing Inaccuracies and Detecting Fraud with Digital Invoicing

App Beats Paper
App Beats Paper

OBJECTIVE

One of our most impactful innovations to date is the ability to automatically generate home-based care invoices based on our caregivers’ GPS data through Helper Hive, The Helper Bees’ mobile app, instead of relying on self-reporting and recollection in the form of timesheets. We have pioneered the use within the long-term care insurance industry of continuous GPS data for individualized geofencing to automatically track, log, and report the number of hours worked by our caregivers. These invoices are generated based on our ability to pinpoint our caregivers’ physical location in relation to our claimants, making it far more difficult to falsely report when compared to submitting timesheets manually. By using continual GPS throughout the visit, The Helper Bees’ mobile app provides carriers with a comprehensive view of the entire visit, not just at clock in and clock out. To date, this is the most accurate and advanced method for generating invoices, detecting fraud and leakage, and is used exclusively by The Helper Bees. However, to validate the superiority of a continual GPS invoicing solution, we wanted to compare our data capture with the alternatives. We are in a unique position to perform this exact data analysis because we assist claimants to file their claims using all current acceptable submission models: fax, telephonic IVR, webform, and mobile-app. This white paper presents data that validates The Helper Bees’ app’s superiority for minimizing soft fraud and leakage. Reducing these losses translates to substantial savings per claimant per year. These savings are monumental when scaled to all enrolled claimants of our long-term insurance carrier partners.

MATERIALS AND METHODS

Two of our insurance carrier partners provide caregivers with the flexibility to submit their hours worked using any method of their choosing. Other than using our mobile app, the most common methods for submitting timesheets are fax, interactive voice response (IVR) via telephone, or through The Helper Bees webform. We reviewed the self-reported hours logged by our caregivers using these common traditional modalities and compared them to the hours logged and reported by our exclusive mobile app. These data were then further validated by ruling out the type of caregiver (agency, informal, or personal) as a confounding variable.

RESULTS

Based on a random sample of 100,000 timesheets, The Helper Bees’ app recorded 6.9 hours per visit compared to the 8.2 hours (118%) per visit on self-reported timesheets. These differences are statistically significant. When compared to the averages: 8.5 hours (123%) for webform, 7.1 hours (102%) facsimile, 7.4 (107%) IVR via telephone, 9.4 (135%) for others, The Helper Bees was consistently statistically significantly lower than all these timesheet modalities. When we evaluated The Helper Bees’ app vs. others broken down by type of caregiver, these findings were validated and reaffirmed. For agency, informal, and personal caregivers, The Helper Bees’ app outperformed everything else: 6.5 vs. 7.6 hours (116%), 6.8 vs. 8.7 hours (128%), and 7.0 vs. 10.3 hours, respectively.

CONCLUSION

These data validate our innovative solution’s superiority to automatically generate invoices since traditional methods rely on self-reporting and recollection, which are inherently inaccurate. Our automatically generated invoices based on continuous GPS data make it significantly harder to fake timesheets and save our insurance carrier partners substantial sums of money. Simultaneously, this innovation also protects our claimants from financial losses and stresses related to overbilling while saving our caregivers time and energy spent recalling, logging, and reporting the number of hours they have worked. These successful, groundbreaking innovations at The Helper Bees beget future successful, groundbreaking innovations.

INTRODUCTION

The Helper Bees distinguishes itself in the long-term care industry by collecting data and translating these data into actionable information through data analytics and machine learning. We have built models that streamline efficiency in all our products, which provided the bandwidth to handle a 9x surge in claimant volume in response to the COVID-19 pandemic. This surge in volume also served to validate the model we use to identify outliers (claimants whose utilization of services is more than two standard deviations away from the mean [or expectations]). In a previous white paper, we published our validation of this model’s ability to both scale and perform independently of temporal variables related to the calendar year, such as open enrollment dates, meeting deductibles, seasonal variations, and more.

In this white paper, we focus on the superiority of modernizing home health and long term care using mobile apps and GPS technology. The Helper Bees has developed a proprietary mobile app, Helper Hive, for our in-person caregivers. The mobile app allows caregivers to submit claimant assessments, digital visit notes, electronic visit verifications, and track services provided. Additionally, the app automatically tracks, logs, and reports the number of hours worked by caregivers, replacing the long-outdated method of self-reported hours using timesheets filled out on paper with pen or pencil.

The Helper Bees’ app automatically logs when caregivers start and stop caring for our claimants. This is done by detecting when caregivers physically arrive on-site and when they leave. This is made possible by continuous GPS location data tracked in and out of a geofence designed for each claimant. This geofence is definable down to the resolution of an individual claimant. By default, the geofence encompasses a predefined radius around the claimants’ home address. The geofences can be adjusted to correct inaccurate address locations and accommodate claimants with homes or lots that are exceeding large. Additional geofences can also be added to individual claimants to account for multiple homes.

The Helper Bees’ mobile app is proprietary and developed in-house, specifically with this task in mind. GPS tracking simplifies and automates the process of logging hours for caregivers. The archaic method of submitting timesheets based on recall and self-reporting is inaccurate for several reasons, but they all amount to leakage and soft fraud. The Helper Bees’ app saves our caregivers time while also automating a cumbersome task that often becomes an afterthought, put off for days or weeks after the care was rendered, which further compounds inaccuracies. It is also preferred by most of our partners because it minimizes financial losses due to padding, leakage, and soft fraud.

This Helper Bees’ mobile app tracks and logs hours worked on-site by our caregivers based on continuous GPS location data mapped to the individualized geofences created for each claimant. It is more accurate and much harder to fake. Continuous GPS location data is more accurate when determining the number of hours worked by caregivers. It results in lower numbers of hours worked compared to traditional methods of submitting timesheets.

This white paper presents data that validates these claims. We compare the average hours per visit recorded by The Helper Bees’ app to timesheets submitted based on self-reporting and recall. We validate our findings by ruling out the caregiver type as a confounding variable. As shown in this paper, our app consistently beats traditional submissions regardless of the type of caregiver. We performed a subset analysis for agency caregivers, informal caregivers, and personal caregivers, and our app consistently outperforms the alternatives. App beats paper.

MATERIALS AND METHODS

The Helper Bees’ long term insurance carrier partners typically require the use of our mobile app (with limited exception), understanding the plethora of benefits, including but not limited to the automated logging and reporting of the number of hours worked through continuous GPS location data. GPS data is more accurate for logging and reporting the number of hours worked compared to self-reporting based on the recollection of our caregivers. As one would expect, self-reporting based on recall tends to artificially inflate the number of hours worked compared to GPS based tracking. This is not necessarily nefarious, like padding the invoice, but could be as benign as fear of under-reporting and therefore being underpaid. However, this could also reflect caregivers’ natural tendency to round up the number of hours worked, including time not directly related to in-person care of claimants or time spent traveling to and from the claimants’ home. These additional hours are referred to as soft fraud or leakage in the long term care industry. In this white paper, we will demonstrate how these costs compound over time. When considering the losses are further compounded across all claimants enrolled with an insurance carrier, it is clear how these losses result in increased premiums for all claimants.

To study the use of continuous GPS location data in relation to geofences, ideally, we could have data from caregivers that use the app and use traditional methods of submitting timesheets. However, these data do not exist. Instead, we designed a study with caregivers from just two of our partners that allow caregivers to choose how they submit their hours. Most of our partners require caregivers to use The Helper Bees’ app unless there are extenuating circumstances. For instance, if there are no cell phone towers near the claimant’s home or the caregiver does not have a mobile phone.

These two partners in this study allow their caregivers to submit timesheets via fax, interactive voice response (IVR) over the telephone, or webform through The Helper Bees website.

In order to share some of our data and methodologies, we elected to sample our databases using 100,000 random timesheets submitted over the last year. This allows us to protect our corporate interests while allowing transparency to substantiate our claims.

To build the data set for this study, we used SQL to query our databases for 100,000 thousand timesheets that met the following criteria. (1) The timesheets were for the two partners that allowed caregivers to freely choose how they wanted to submit the number of hours they worked. (2) The claimants had to still be living at the time of the query. (3) The claimants had to have at least 30 timesheets (decrease random variance). (4) The claimants do not have any cognitive needs (for similar reasons in previous papers, these claimants represent a minority of claims and introduce noise into our model. (5) The timesheet was submitted within one year (Sept 4, 2019 — Sept 4, 2020 [366 days due to leap year]). (6) At least 30 minutes was logged or reported for the visit (this decreased the chance of canceled appointments affecting the cohorts). (7) Twenty-four or fewer hours were logged or reported for the visit (these are screened out during data collection and entry).

The data set was complete; therefore, no exclusions were applied.

RESULTS

Figure 1 visualizes the distribution of the 100,000 randomly selected timesheets based on the modality of submission. The most common modalities outside of The Helper Bees’ app, webform, faxed, and IVR were consolidated into “Other.” Remember, these timesheets were chosen from the insurance carriers that let their caregivers freely decide how to submit their hours worked per visit. When given a choice, caregivers most often choose The Helper Bee’ app (39.8%), representing more than a third of all submissions. The second most common method is submissions through The Helper Bees’ webform (27.1%), accounting for more than a quarter of all submissions. Finally, more traditional methods of submitting via fax (15.9%) and telephone via IVR (5.7%) together consist of about one-fifth of all submissions.

Figure 1: Distribution of 100,000 timesheets based on the method of submission.

Figure 2 visually demonstrates the average number of hours tracked and recorded using The Helper Bees’ app versus all other submission modalities. On average, The Helper Bees’ app tracked and logged 6.9 hours per visit across 39,782 visits (~40% of all submissions). Timesheets submitted using all other modalities averaged 8.2 hours per visit across 60,218 visits (~60% of all submissions). When comparing the app to paper (non-app), non-app submissions are 118% higher on average, reflecting a substantial amount of “padding the timesheets.”

Figure 2: The average number of hours per visit based on The Helper Bees’ app submissions versus all submissions types.

Figure 2: The average number of hours per visit based on The Helper Bees’ app submissions versus all submissions types
Figure 2: The average number of hours per visit based on The Helper Bees’ app submissions versus all submissions types

Figure 3 depicts the difference between hours per visit as tracked and recorded using The Helper Bees’ app compared to caregivers submitting their hours through the webform (through The Helper Bees website), based on self-reporting and recollection. These data show the number of hours per timesheet or visit using the webform is 8.5 hours (27,100 timesheets, about 27% of all timesheets). This is 123% higher than the average hours per visit as logged using The Helper Bees’ app.

Figure 3: The average number of hours per visit based on The Helper Bees’ app submissions versus submissions through the webform.

Figure 3: The average number of hours per visit based on The Helper Bees’ app submissions vs submissions through the webform
Figure 3: The average number of hours per visit based on The Helper Bees’ app submissions vs submissions through the webform

Figure 4 depicts the difference between hours per visit as tracked and recorded using The Helper Bees’ app compared to caregivers submitting their hours through facsimile, which is similar to webform in that it is based on self-reporting and recollection. These data show the number of hours per timesheet or visit using facsimile is 7.1 hours. This is 102% higher than the average hours per visit as logged using The Helper Bees App.

Figure 4: The average number of hours per visit based on The Helper Bees’ app submissions versus submissions through facsimile.

Figure 4: The average number of hours per visit based on The Helper Bees’ app submissions vs submissions through facsimile
Figure 4: The average number of hours per visit based on The Helper Bees’ app submissions vs submissions through facsimile

The data shown in Figure 5 for IVR is interesting because caregivers’ adoption rate has almost tripled in the last year averaging ~50 per week and growing to ~150 per week. This is depicted using the dotted yellow line. The increased variation in the weekly IVR data in the first half of the graph is secondary to the low sample sizes. Overall the use of IVR is low, representing only 5.7% of all timesheets (figure 1), but as its use increases, the difference is more and more clear. The Helper Bees’ app records an average of 6.9 hours per visit, compared to 8.0 hours per visit when caregivers self-report via IVR in the last six months.

Figure 5: The average number of hours per visit based on The Helper Bees App submissions versus submissions through the telephone using IVR.

Figure 5: The average number of hours per visit based on The Helper Bees App submissions vs telephone submissions through IVR
Figure 5: The average number of hours per visit based on The Helper Bees App submissions vs telephone submissions through IVR

Table 1: Average hours per visit based on the type of submission.

Table 1: Average hours per visit based on the type of submission
Table 1: Average hours per visit based on the type of submission

Figure 6 shows the distribution of timesheets by the assigned caregiver type. These sample sizes do not add up to 100,000 because the caregiver type is not disclosed by one of the two carriers that allow caregivers to choose recording methods other than The Helper Bees App. Over half of the caregivers (53.2%) are contracted through an agency (53.2%). The “personal” caregivers are unique to California. These “personal” caregivers are a subset of “informal” caregivers that are granted special designations in order to receive higher reimbursement rates.

Figure 6: The Distribution of caregivers based on their category (1) agency, (2) personal/private, and (3) informal/family member.

Figure 6: The Distribution of caregivers based on their category (1) agency, (2) personal/private, and (3) informal/family
Figure 6: The Distribution of caregivers based on their category (1) agency, (2) personal/private, and (3) informal/family

Figures 7–9 demonstrate the consistent ability for The Helper Bees’ app to outperform all the other submission types irrespective of the type of caregiver. The differences in means are all statistically significant. These data are summarized in Table 2.

Figure 7: Average hours per visit submitted by Agency/Home caregivers comparing The Helper Bees’ app to all other submission types.

Figure 7. Average hours per visit submitted by Agency/Home caregivers comparing The Helper Bees’ app to all other submission
Figure 7. Average hours per visit submitted by Agency/Home caregivers comparing The Helper Bees’ app to all other submission

Figure 8: Average hours per visit submitted by Informal caregivers comparing The Helper Bees’ app to all other submission types.

Image for post
Image for post

Figure 9: Average hours per visit submitted by Personal caregivers comparing The Helper Bees’ app to all other submission types.

Image for post
Image for post

Table 2: Summary of The Helper Bees’ app versus other submission types.

Image for post
Image for post

DISCUSSION

Typically, our partners do not require caregivers to use our mobile app; instead, they offer alternatives should the claimant prefer a different submission method. (Caregivers are not influenced by external factors such as reimbursement rates.) This creates the perfect opportunity to validate the superiority of The Helper Bees’ app with data since most partners intuitively expect our mobile app to outperform the outdated methods of submitting timesheets based on self-reporting and recollection of caregivers.

The app tracks and logs the hours caregivers spend delivering care to claimants. The hours of service per visit are based on continuous GPS data in relation to geofences established for every claimant. When caregivers spend too much time outside of these geofences, they must provide a valid reason before they are compensated for their time outside of the geofence. These reasons can include but are not limited to assisting a claimant with shopping or errands. The continuous GPS data is much harder to falsify than traditional methods of reporting the number of hours worked.

Not only does The Helper Bees’ app more accurately record the number of hours worked, but it also saves caregivers a substantial amount of time. Caregivers no longer have to attempt to recall the number of hours they spend providing care to their claimants. When caregivers are faxing their timesheets, there is a significant delay between when care is rendered and when the timesheet is faxed. From these same data, a month-over-month analysis of faxed timesheets reveals ~20% of fax submissions are still outstanding and haven’t been turned in. These substantial delays further contribute to the inaccuracies of timesheets based on self-reporting and recollection. For all of these reasons and more, 40% of caregivers prefer and voluntarily use our app to track their hours of care per visit.

The Helper Bees’ app is preferred by our caregivers and insurance carriers alike. In fact, most of our partners require their caregivers to use the app barring extenuating circumstances. The reason for their preference is clear. Figure 2 shows how our app consistently outperforms all other methods of self-reporting the number of hours worked per visit. The differences are substantial. All other forms of reporting hours worked are effectively padded an additional 18%. This leakage or soft fraud is substantial and affects the bottom line. These losses are even more pronounced when comparing The Helper Bees’ app to webform submissions. Keep in mind, the app does not rely on self-reporting because everything is tracked and logged automatically using continuous GPS location data. The webform is the most common method caregivers use to submit timesheets based on self-reporting and recollection. Similar to Figure 2, Figure 4 shows how the app outperforms the webform every week over the last year. However, these differences are even more substantial, averaging 23% more hours per visit for webform versus The Helper Bees’ app. Table 1 shows how The Helper Bees’ app outperforms all other submission types. All of these differences are statistically significant.

To further validate these claims, we wanted to ensure there were no confounding variables. The most likely and most important confounding variable to rule out would be an unforeseen association with the type of caregiver. More than half of our caregivers are contracted through a home health agency (~53.2%). Informal caregivers are typically family members who take care of their loved ones and submit claims to get reimbursed for their time and services. Personal caregivers are informal caregivers that are granted a special status only in California that qualifies them for higher reimbursement rates.

Given the differences in the types of caregivers, we performed a subset analysis of The Helper Bees’ app versus all other methods of submission based on each of the three types of caregivers. These data are presented in Figures 7–9 and summarized in Table 2. These figures show that the app consistently outperforms the use of timesheets to submit the number of hours worked per visit. Even when comparing the data from personal caregivers, where the variation is higher secondary to the smaller sample size, the app still outperforms using timesheets submitted through alternative methods.

These data present clear evidence. The Helper Bees’ app consistently tracks and logs fewer hours per visit than methods that rely on self-reporting based on recollection. This is consistently seen temporally (weekly over the last year). This is also consistently seen when comparing different submission modalities such as facsimile, IVR via telephones, webform, and others. This is also consistently seen when considering all the types of caregivers: agency, informal, and personal.

Currently, there is no gold standard for logging and reporting hours worked by caregivers. However, we propose using our mobile app-based GPS location data and geofencing as the standard because it accurately and automatically tracks when the caregiver is physically located near their claimants’ home(s). This method is not perfect and does not completely eliminate fraud, but it is a far superior method for logging and reporting the number of hours worked per visit versus traditional methods that rely on self-reporting and recollection.

Before we file a claim with our partners, we first generate an invoice based on the caregiver’s timesheets. This invoice has to be reviewed and signed off on by the claimant. Historically this served as the only check against soft fraud and leakage. However, these crucial checks are cumbersome, time-intensive, and relied upon claimants who are rarely incentivized to review and contest timesheets submitted by their caregivers.

The Helper Bees’ app changes this process drastically while preventing soft fraud and leakage in two very important ways. First, as we have proven in the paper, The Helper Bees’ app tracks and logs fewer hours per visit than methods based on self-reporting and recollection. Second, the app makes it significantly easier for claimants to contest timesheets. Internally, we see claimants flagging about 10% of invoices for review when using The Helper Bees’ app. Conversely, claimants flag less than 1% of invoices for review when using other modalities finding the process too time-consuming and cumbersome.

When we consider the cost savings of app-generated hours per visit versus timesheets based on recollection and self-reporting, these small differences in hours per visit translate into large differences in expenses for our insurance carrier partners. We calculated the average costs savings based on the difference in hours per visit reported using the app versus non-app methods. Figure 10 depicts these differences when multiplied by the average hourly rate across the nation for in-home caregivers at $21/hour. These differences are more apparent weekly and even more striking, considering the average cost savings over a year. The potential savings for insurance carriers when considering all of their claimants is staggering.

Figure 10. Comparing costs per claimant per year

Image for post
Image for post

CONCLUSION

The Helpers Bees prides itself on being the ultimate partner for both our claimants and insurance carriers. We have transformed the long-term care industry by giving claimants what they want: a system built around the desire to age in place. Care is individualized and tailored to the needs of our individual claimants. At the same time, we increase revenues and decrease expenses for our insurance carrier partners. The Helper Bees is able to enhance the experience on both sides of the relationship between insurance providers and claimants. This transforms the traditional tug-of-war between people and insurance companies into a symbiotic relationship since The Helpers Bees aligns everyone’s incentives in the same direction.

The Helper Bees’ app is an extension of these core values, helping caregivers, claimants, and insurance carriers alike. The app reduces caregivers’ stress and time commitment while recalling and self-reporting the number of hours worked each visit and submitting them through fax, IVR via telephone, filling out a webform, etc. The preference for using The Helper Bees’ app is reflected in the 40% of caregivers who preferentially elect to use our mobile app. Similarly, the app is preferred by claimants because they know hours of care they receive are being tracked and logged automatically. They do not have to worry about caregivers padding the timesheets and incurring undue costs. Finally, most of our partners prefer their caregivers use The Helper Bees’ app because of the substantial cost savings and an ROI of 11x annually (Figure 10).

In addition to reporting the number of hours worked, our caregivers also submit assessments of our claimants’ needs and the intensity of the services they provided to the claimant. These data points are critical in developing our model that predicts our clients’ needs, which are discussed in a previous white paper regarding our predictive care model. These data points are also used to build models that help predict and highlight potential fraud. The Helper Bees’ app is also monumental in that it automatically scrubs the data and keeps our databases clean and ready for on-demand data analytics. Therefore, our goal is to steadily increase the share of caregivers and partners using our mobile app.

The ability of The Helper Bees to generate products that serve in the best interests of our claimants, caregivers, and insurance company partners is what makes us a leader in insurance technology, long term care, and home health. We continue advancing and modernizing the status quo of an outdated industry. We strive to be a leader in innovation to serve claimants, caregivers, and partners while staying true to our core values and mission to build systems around individuals so they can happily and safely age in place.

Jitsen Chang, MD

Medical Data Scientist

Jessica Faulk

Data Scientist

Char Hu, PhD

Chief Executive Officer

Written by

#Insurtech #AgingTechnology #InHomeCare #Thinker CEO of The Helper Bees @thehelperbees and founder of Georgetown Living @georgetownalf

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store