How do we enable doctors to adopt tech (and Bahmni in specific) in LMIC?

We are seeing low adoption of Bahmni by doctors in one of the (pilot) deployments that we have done for one of the states in India. It is deployed at a higher health facility where the patient load is high. As a part of change mgmt, extensive training had been provided to them, there is a direction from the state as well to adopt the platform. But the adoption is still low, especially with doctors and not with other users.

Our assumption is that it can be due to any of the following reasons:

  1. They do not want an overhead of entering data – reluctance to adopt tech
  2. High workload
  3. Additional responsibility with no linked benefits

A possible solution to this issue was adding a data entry operator along with a doctor, but that would be an extra cost to the state and might not be sustainable in the future. Hence, this does not look like the approach that the state might be ready to take up at the outset.

Another possible solution using technology, can be

  1. A speech to text converter a. Can we use Windows or GBoard speech to text capabilities with Bahmni?
  2. A digital pad that can convert hand writing to text Hence, I have the following queries:
  3. Is there something like this that has been integrated with Bahmni in the past? Or is there any other tech that can be used OOTB with Bahmni to address such issues?
  4. Also, from your experience with LMIC rollouts so far, what else has worked out for increasing adoption of Bahmni (and tech platforms in general) with doctors ?

Looking for solutions across all the change management heads - people, process or technology.

Thanks in advance!

2 Likes

Hi @maheshwari.neha03,

This is a great question, and a frequent challenge in most EMR rollouts across the world (not just for Bahmni) — i.e. how do we increase EMR adoption by clinicians, without impacting patient care – or maybe just – how do we make more clinicians agree to using an EMR.

There seems to be a lot of research material on the internet on this topic. And I think, the solution would likely be a mix of: Operational, Management, Motivation, Capability Development, Expertise, Time, Feedback, etc… and also some EMR-centric feature development. We need to be careful to think that building a “new feature” will “increase adoption”. It is unlikely the main root cause.

For instance this paper mentions many reasons for low adoption of EMRs by Physicians: The adoption of electronic medical record by physicians.

My suggestion would be to dig deeper, and truly see why clinicians are hesitant in adoption bahmni. In this case the “solution lies in the problem”, and digging deeper into the problem will help.

I would also recommend reading this research paper by Possible Health (@anant) describing their rollout strategy for Bahmni in Nepal. Link: Design and implementation of an affordable, public sector electronic medical record in rural Nepal

Some things they tried that worked:

  1. Institutional (leadership) buy-in into wanting to use an EMR and helping staff see the advantages.
  2. Phase wise rollout, to allow for feedback, minor steps, getting comfortable and buy-in.
  3. Training, patience, evangelizing and champions.
  4. Button based interface for quick form entry (on tablet) + chromebooks.
  5. Small focus groups + regular feedback + incorporating the feedback for building a sense of ownership and pride.
  6. Metrics that can show adoption and some impact.
  7. Ensuring system uptime to build confidence.

In LMICs, clinician time/bandwidth is premium. It is possible they are already so busy, that they don’t even have time to write paper reports in most cases – and hence asking them to type into a computer system might be impossible. But, i feel any change usually needs to be supplemented with right incentives to overcome the effort of incorporating it (see beckhard-harris-formula-for-change). And, we need to investigate both sides – the effort, and the incentives.

Meanwhile, Bahmni team is looking into OCR - for rapid Lab report uploads. But that will still only solve a tiny part of the problem. Speech to text solution also sounds promising, but the tech and real-world constraints still need to be worked through. Its feasibility in LMICs seems low, but it’s improving.

I would love to hear inputs from other folks. cc: @angshuonline @akhilmalhotra @arjun @vmalini @michaelbontyes @mksrom #openmrs #pih @MekomSolutions @burke @grace @anant

2 Likes

A related post (and demo) on Speech to text by @ramashish: Bahmni with Dictation module

1 Like

Gurpreet perfectly summarized it. The lesser the work, the better the adoption. That’s why a lot of eCDSS apps have a good adoption rate because it guides them throughout the care. They don’t have to look at the guidelines, the app tells them what to do based on their input. In my experience (from different organizations), the main challenges for having low adoption are caused by the following:

  • High volume of patients (emergency cases) - not enough clinicians for the number of patients coming in.
  • Staff turn over - we can’t keep up with the training (if a replacement is coming in) or if there is no replacement, then there is a staff shortage
  • Added value - There is no benefit in using the tool (for them). They see the tools as a tool for the organization for reporting only.
  • Support - when system goes down (or any issues - e.g. slowness), they need someone to be available immediately. They also need someone to enter the information that was not entered during the downtime. If the patient comes back and the information was not entered, and they don’t have the capacity to enter data themselves due to high workload, then this might cause them to go back to the old solution that has worked for them for so long. This also involves training requests and training materials available.
  • Expectations - we expect everyone to perform the same way as they did (or better) before the introduction of the tool. But everyone has different learning style (but the training delivery is the same across). Some people adapt faster than others

Some of the factors that helped in facilitating the change within a facility:

  • Ownership - they need to own the information, and know how to get the information they need (and have the support they need when they need it).
  • Feedback Loop - This ties into ownership - involve the doctors when designing the application. They have to feel that their voice is heard and they make a difference
  • Bottom up approach - The initiative has to be started by the people in the field, the clinicians. They have to define the problem and come up with a solution. Even though you already have a solution, they can always bring something to the table to improve the tool.
  • Data Analysis - resources are put in place to ensure data is clean and reliable. Have data in near real-time. Something different from the information produced using the traditional method of data collection/reporting (e.g. total number of monthly consultations by age/sex). The tool has to bring something new that will allow them to improve the quality of care or see an opportunity for improving existing practices. Show diseases outside of the normal data the ministry is asking for. For example, showing that the number of ER visits are higher during a specific period of time (this will allow the decision makers to plan more staff during these hours to avoid long wait time).
  • Executive Leadership - The organizational leadership has to be fully on board and have a policy on the tool being implemented. If it’s a mandatory tool, say it’s mandatory but provide all of the necessary resources for the team to be sucessful. If all the resources are put in place, and they’re still not in compliance, then what’s going to happen? If nothing happens, then why change the existing practice. If they do comply, then what’s the benefit. Managers to keep track and ensure the team are complying, if not, then understand why.
  • Culture - we just have to accept that there are people that are completely against everything no matter what and know who they are and be there friend. Understanding why they resist to change and understand how to make them your champion can really make a big difference. This one person can make your deployment a success or a failure.
  • Champion - Tied to the above, find a champion at all levels of the organization and make this person promote and support the tool. Some hospitals in the U.S. (at least in the beginning of digitalization) have introduced a position called “Physician Liason” who basically is the one person managing the resisitance from doctors (especially those who are retiring).

As far as adding additional tool to make the data entry easier, this can help but may introduce another set of issues. Speech recognition or OCR really depends from one person to another. It may work for one doctor but it doesn’t for another. The words are not recognized by the system (either speech or OCR) and will lead to frustration.

5 Likes

Great insights from Gurpreet and John. I totally agree with them. This is the problem EMRs face not only in India but across the world.

Apart from the reasons mentioned above, in my experience, other reasons for reluctance in adoption in govt hospitals:

  1. There is no tangible benefit for the doctors especially in district hospitals since the doctors in these facilities mostly cater to a mixed general OPD (~80%) i.e. they see all kinds of patients every day and these patients seldom come back for a follow-up.

  2. There is no motivation for a pro-active patient follow-up for chronic conditions unless the patient is admitted.

  3. Time is at a premium. A physician hardly gets 2 minutes per patient due to the huge patient load.

Technology solutions like speech to text and OCR will indeed lower the barrier to adoption but these will only work if there is enough motivation in the care providers to use the system.

My suggestion would be to have a two-pronged approach to improve adoption.

  1. Bottom-up:
  • Rather than doctors, aim to increase adoption by the support staff especially nurses. District Hospitals typically have at least 50 to up to 100 beds. I would suggest pushing for adoption in IPD first. Doctors in the govt sector often delegate writing the discharge summaries to the IPD nurses and just review and sign them.

  • Bahmni can automatically create discharge summaries that can be just printed and handed over, thus reducing the workload of the nurses. That could be a starting point to showcase the capability of the system to the nurses and doctors.

  1. Top-down: Data is something that excites the hospital leadership as well as the district officials as John mentioned in his points on Data Analysis and Executive Leadership.

If both the approaches work in conjunction, a push towards adoption and at the same time showcasing the benefits at the ground-level would help adoption at least in certain care settings like IPD and chronic care programs.

2 Likes

@gsluthra @jesplana @akhilmalhotra great thoughts and points.

@maheshwari.neha03 , I can completely relate to this. Imo, any EMR implementation, where a doctor gets minimal time is a challenge for adoption.

You can still do a lot with thoughtful planning, devise a phased approach with consideration to all important aspects, as John and Gurpreet mentioned. For HIMS implementations, where I have seen some success, I do believe thoughtful roll outs in phases with value based plans have worked well.

Implementations, such as yours, are also about shifting mindset towards using data for purposes, building such culture takes time. We cannot expect users to adopt and adapt without incentives, without them deriving values. In India and similar contexts, probably for a doctor (in public health setup) an EMR is clinically useful only for Chronic complex cases. While many routine information elements may be extremely useful, orgs who judiciously record and utilise such info are more driven by research and advocacy purposes.

One aspect, we don’t think/design deeply about is data minimisation. Can we design/roll-out a solution without information overload, that’s contextual for the user, and emphasise on data minimisation? I have seen many implementations designed towards collecting medico-legal, billing information, M&E information. Often, such piles of data are almost never analysed. For adoption, we need human centred design without providers as data-entry operators.

Somethings that have worked while devising such phased rollouts (delivering incremental value):

  1. start with profiles of user that take to HIMS early - registration, lab/pharmacy, radiology, Vaccination centers - as they immediately realise the values/benefits.

  2. Doctors to only capture provisional diagnosis, medications and orders to start with. eg., Can information be sourced and correlated from other services points? (e.g. in Bahmni, even without doctors orders, when tests are done at Labs - the details flow to EMR).

  • Only in services where protocols are well established (e.g a HIV/MCH/NCD) - mandate observations to be captured, symptoms/allergies to be recorded. In all other cases, trust your doctor to record relevant information at her discretion.
  • One time assessments - like cancer screening, diabetes screening etc

Implementations often don’t design/accomodate for evolution and puts lesser effort on change management and adjustments. Most implementations go with a template, a set plan. You would fare well to adopt an iterative approach of figuring out what works in a given context.

1 Like

Thank you @gsluthra, Akhil, Angshu and @jesplana for sharing your thoughts and such relevant insights, research papers, and valuable suggestions. The Chrome speech dictation is definitely worth trying but would be really difficult in the Indian scenario considering the language and dialect changes every 50-100 km.

As of now, in the current implementation, the doctors are being assisted by a data entry operator - not the best way to approach the problem, given it is not scalable and there are bound to be data entry errors.

We are also trying to hear from the doctors on why they are resisting this adoption - is it workload (COVID or otherwise), low incentives, or something else.

Would keep this group posted on our findings! Thanks again! :slight_smile:

1 Like

A short youtube-demo of using Voice to enter consultation notes in Bahmni. This was created recently by a team in Thoughtworks that integrated Bahmni with Vakyansh (Indic speech-to-text engine that supports many regional languages).

Note: This is just a prototype activity, and therefore not in the Bahmni core codebase yet. Follow this page for updates on this initiative (point #10).