Category: Digital Health

One of healthcare’s biggest venture capital hot streaks is adding fuel to the fire, with General Catalyst and WellSpan Health entering a new partnership to drive digital transformation.

General Catalyst and Pennsylvania-based WellSpan are aiming to improve care models and patient engagement by embedding tech from the VC firm’s “health assurance network” into the health system’s operations.

GC’s health assurance network is comprised of portfolio companies with broad expertise in everything from employer benefits to population health, and includes marquee names such as Aidoc, Cadence, Olive, Sword Health, and Transcarent.

• The partnership involves no financial commitment from either party, but provides GC with insights from real-world clinical applications and allows WellSpan to invest in any co-developed solutions.

• The formula appears to be working. This is GC’s fourth health system partnership of this kind, adding to collaborations with Intermountain Healthcare, Jefferson Health, and HCA Healthcare.

General Catalyst originally found success with grandslam investments in Airbnb, Instacart, and Snapchat, but it’s been pretty hard to miss the huge waves it’s been making with its dive into healthcare.

• In July, GC raised its second $600M+ healthcare fund, which it followed up by tapping then-Intermountain Healthcare CEO Marc Harrison to helm its investment platform.

• That also wasn’t exactly GC’s first hospital exec hire, with former Jefferson Health CEO Stephen Klasko joining the firm in February… Looking back up to the names of GC’s health system partners, WellSpan might want to take a hard look at its incentive packages if it wants to hold onto its leadership.

The Takeaway

General Catalyst couldn’t make its stance on healthcare transformation any more clear: incumbents hold the keys to the innovation kingdom. The VC firm is also keen on catalyzing its own thesis, so expect plenty more health system partnerships to be inked before the end of the year.

Patients are more likely to report a positive telehealth experience than providers, at least according to Zocdoc’s new Healthcare Experience Report that broke down many of the highs and lows of delivering virtual care over the past two years.

Zocdoc analyzed the trends from its appointment booking platform between May 2020 and May 2022, then conducted patient and provider surveys to uncover insights into how care experiences evolved throughout the pandemic. The final report got pretty granular with different specialties and care types, but delivered a few overarching themes worth taking a look at.

If you’re more of a visual learner, Zocdoc also put together an infographic that does a good job summarizing the main findings.

Differing perspectives between patients and providers are common. Just over 30% of patients surveyed indicated it was easier to build a relationship with their provider via telehealth due to less formality and more relatability in the interactions. Only 7% of providers felt the same way.

• Similar discrepancies were seen with tech difficulties. While 58% of providers indicated they or their patient has had tech issues during an appointment, just 30% of patients said they’d ever experienced a problem.

Mental health is the only specialty where virtual care remains dominant. No surprise here, but it was interesting to see how virtual mental health appointments have continued to overtake in-person visits over the last couple of years.

• Virtual appointments comprised 74% of mental health bookings in May 2020, 85% in May 2021, and now account for 87% of appointments. Survey respondents cited convenience and the benefits of at-home comforts as reasons they appreciate virtual mental health visits.

Telehealth is a supplement, not a substitute. In May 2020, one-third of all appointments booked via Zocdoc were telehealth visits. By May 2022, that number had declined to 17%, and it dwindled to just 9% if you exclude mental health appointments.

• Although telehealth is well above its 1% pre-pandemic numbers, 77% of patients reported that they’re still leaning towards a combination of in-person / hybrid care. This is easily seen in the rebooking trends, where over half of virtual visits with a new OB-GYN, eye doctor, or dentist resulted in an in-person follow-up. 

The Takeaway

We’ve seen many of the themes from Zocdoc’s report before, but it’s nice to wrap some numbers around them from an appointment booking perspective as opposed to the usual claim lines or anecdotes. The discrepancies between patient and provider views on telehealth were also good additions that get studied less frequently, as were other hidden gems like the fact that 36% of providers have seen a patient’s pet during a video appointment.

Risk scores are used to help predict everything from hospital readmissions to medication adherence, but more-often-than-not these valuable metrics are still generated by adding up a few key predictors in a simple model. That makes the difficult task of selecting the right variables extremely important, which is why researchers out of Duke-NUS Medical School in Singapore developed a machine learning system to help optimize the process.

Most current risk scores are generated from regression models that are popular for their simplicity and predictive power. The widely-used LACE index predicts unplanned hospital readmissions using only four components: inpatient length of stay, acute admission, number of ED visits in the past 6 months, and comorbidities.

There have been several efforts to improve risk scores using advances in AI, but these models are usually “black boxes,” making it notoriously difficult to interpret variable selection and importance.

The Duke-NUS researchers proposed an AutoScore-ShapleyVIC machine learning model for improving risk score variable selection while maintaining interpretability, then compared it to the LACE index for predicting 30-day readmissions of patients who visited the ED of Singapore General Hospital between 2009 and 2017.

We’ll leave the model development details to the AI experts, but here’s a look at how AutoScore-ShapleyVIC performed against the LACE index:

• AUC: 0.756 (AutoScore-ShapleyVIC) vs.0.733 (LACE)
• Accuracy: 71% vs. 64%
• Sensitivity: 66% vs. 74%
• Specificity: 72% vs. 62%

The final AutoScore-ShapleyVIC model significantly outperformed the LACE index, but didn’t sacrifice interpretability to achieve its results. Although it was able to narrow 41 candidate variables down to just 6 for the final model (number of ED visits in past 6 months, metastatic cancer, age, sodium, renal disease, and ED triage), the logic behind the variable selection can still be visualized to help with interpretation.

The Takeaway 

Although we don’t usually cover AI studies, this research helped illustrate that machine learning models have a lot of potential to improve existing care paths without sacrificing interpretability. The authors emphasized how this particular approach is not limited to any specific clinical application, suggesting that machine learning algorithms are still in the early stages of improving risk scores in other areas such as ED triage or cardiac arrest survival.