Organisational fragmentation

Participants worked across separate, fragmented health and care departments and organisations such as CCGs and local authorities. Those who recounted facing challenges when they had attempted to use analytics described how divisions between, and within, organisations created siloed data systems, which meant residents’ records could be stored in different data systems if they contacted more than one service. At times, this made it difficult for leaders to access data as they had to actively request data and/or analytics from individuals in other departments or organisations. Divisions in systems across organisations meant senior leaders did not always know who held certain data, whether the data they held would be relevant to inform decision-making or how to contact key individuals. These barriers were aptly described by one participant who had tried to access analytics to better understand and plan for social care accommodation needs:

[RES]: We need housing data, we need social care data, we need some health data, but it’s proving difficult to get those data sources … there’s issues around [asking] “where does the data sit?”. So, I had a meeting with [an internal team] asking for some data. They’re like, “But this sits here, this doesn’t sit with us”. It’s unclear who owns certain pieces of data and how best to extract it.

[INT]: Is that the reason that you had issues accessing it in the first place?

[RES]: Definitely. So, housing data, in particular, where it sits [is] in a completely different department, a different team. We have no right to access any of that data, so it will take quite a lot of time to get it. (ID023, social care commissioner)

For some, information governance requirements contributed to these barriers to data sharing across departmental and organisational boundaries. For example, General Data Protection Regulation, which is a legal framework for the collection and processing of personal data introduced in the UK in 2018, was described by one participant as follows:

GDPR was supposed to make [sharing data safely] better or easier but I think that’s caused a lot of complications as well … I think the trust [for me] is, [once I’ve shared my data with you] are you going to ensure that you’re following the rules, so if there’s a breach, I don't have to pay ten percent of my revenue. (ID015, health commissioner)

Overcoming barriers to data sharing often involved a time-consuming process, where participants had to identify who to request data from and justify their need. The former participant continued:

Having to explain the rationale as to why we need data is always the start of it and can always be a bit of a challenge [in] trying to make them understand why I need access to this data and what it will be used for … . But I think the biggest thing is, everyone’s busy … . it’s never a priority when someone else comes saying, “Do you have this data source? I need it for X”, because I think, “I’ve got twenty other things on my plate”. (ID023, social care commissioner)

This time-consuming process requiring continuous justification was, therefore, described as an additional organisational barrier to data sharing, analytics access and use of analytics—with other priorities and work often taking precedence. When participants could not access data held on siloed systems, some made decisions without all the ‘necessary information’ (ID022, health and social care commissioner), while others relied more on expert opinion (such as the opinion of single practitioners) or stopped their use of analytics.

Competing priorities

Many participants described how fragmentation across their health and care system, at times, led to different or competing organisational priorities. In more extreme circumstances, this hindered collective priority setting for health and care decisions, despite organisations being encouraged to align priorities locally to facilitate collaboration. For instance, one participant expressed little motivation to engage in health and care decision-making and promote data sharing due to conflicting financial drivers:

If we have a patient who we see in the hospital we get paid £70 or something for a follow-up. If we work out a new model of care where this patient can be seen in the community or virtually, we would get paid £10 or £15. What on earth would we want to do that for? … If you’re saying let’s [in a] wholesale [manner] move half of our patients into the community, let’s lose all of that revenue, then suddenly the fixed costs that we have in this building and others become overwhelming. (ID011, health provider)

Indeed, several participants described how reservations around sharing data often stemmed from conflicting priorities. In addition, some participants stated they were more likely to share their data if they trusted that recipients had priorities aligned to their own and, as such, would use their data as they had specified. This was particularly relevant for data sharing between commissioners and providers, where providers were hesitant to share data in case commissioners used it to justify disinvestment in their services.

Interviewees also observed that they were often competing for analysts’ time against the extensive mandatory requirements they faced from external public bodies such as NHS England:

The structures that sit across us, there are data requirements placed on us which are often at short notice and unexpected or slightly different or very similar to one that we did previously. The time and energy and resources that it takes for [analysts] to keep changing that information and updating it and translating it into the latest format is time consuming, it’s energy sapping … So, yeah it’s not [the analysts’] priority to respond to our [analytics] requests immediately. (ID022, health and social care commissioner)

Externally mandated requirements that occurred frequently, unexpectedly and at short notice were, therefore, described as creating ‘time consuming, energy sapping’ work that needed to be prioritised over requests from leaders for analytics support. This was described as a barrier to analytics access, which hindered leaders’ use of analytics.

Leader–analyst relationships

Participants suggested that the uncoordinated way in which leaders obtained analytics meant relationships between leaders and analysts greatly influenced analytics access and use. Some leaders who experienced advanced use of analytics regularly collaborated with trusted analysts to obtain suitable analytics support. They described having a ‘good dialogue’ with analysts, which facilitated data access, and enabled leaders to iteratively and successfully review and use analytics to inform decisions. In explaining how a collaborative relationship with an analyst worked, one participant said:

We kind of described the scope of the strategy, and what we'd intended it to do, and then [the analyst] went off and led [the work]. We had a couple of meetings to check in every so often … [the analyst] and I have worked together on and off for years … I just inherently trust [the analyst] to know what [they’re] doing. (ID020, health commissioner)

The benefits of having ‘good’ working relationships with analysts appeared so crucial that leaders ‘attach[ed] themselves to good analysts, even if external to their organisation:

There’s a better analyst in [an external organisation]. [And so,] I would nick [them] sometimes. I would trust [their] judgement around [how the analysis should be conducted]. (ID021, health and social care commissioner)

Participants who faced barriers when trying to make analytics-informed decisions typically stated that, while they wanted collaborative working relationships with analysts, these were not currently available. In some cases, some of these participants could not access data as they did not know who to contact. Those who could access data, but were reluctant to use analytics, described a struggle to develop questions that could be addressed without analysts’ input. This led to ‘insufficient’ outputs, which did not address questions they required answering, lacked extra detail around how to interpret and use the output, or recommended unfeasible actions.

Organisational fragmentation was also described as creating physical disconnect between leaders and analysts, meaning that good, cross-organisational relationships were even more salient. For instance, one provider faced difficulties working with external analysts, as outputs did not contain details necessary for their decision-making. They felt this was because analysts were not ‘part of the team’ and, therefore ‘didn’t know what [the leaders were] talking about and leading on’ with respect to a decision (ID012, health provider). This participant eventually hired an internal analyst to produce better-suited analytical support. Several participants believed that they had a better understanding of how services operated than analysts because analysts where not co-located in decision-making teams. This drove their choice to request raw data and conduct their own analyses to support decision-making, independent of analysts’ input.

Leader–leader relationships

Building trust and relationships between individual leaders of the organisations was also vital for some participants when making strategic decisions across organisational boundaries. More regular and confident users of analytics had established relationships and aligned strategic priorities with other, trusted leaders. Conversely, those who faced barriers to obtaining and using analytics from external organisations typically faced difficulties forming relationships with other leaders and aligning strategic priorities:

We’ve got a new Director [of the partner organisation] come in, who very much sees that they’ve got to sort out this little corner [of the decision] as a separate project, rather than doing it all at once. Which has delayed the togetherness of the project.…We were talking 18 months ago, we'd got the model ready, and yet we’re still sitting here now, talking about it. (ID016, health provider)

High turnover of senior leaders, in general, was also described as a barrier to developing and sustaining leader relationships, stalling project delivery and use of analytics.

Data availability and accuracy

Several participants described circumstances where data they required for a decision did not exist because certain groups had little or sporadic contact with services or recording of certain information was not mandatory. This hindered their ability to make decisions for these populations. For example, when discussing service provision for residents with autism, one commissioner stated that they ‘simply don’t know how many children have autism, because there are whole cohorts not recorded’ (ID022, health and social care commissioner). They went on to describe how this made it difficult to accurately plan services, as they could not determine how may children had autism in the borough. Attempting to overcome this issue, they retrospectively collected data, which was a resource-intensive and ‘frustrating’ task. They also relied on ‘professional judgement’, ‘gut feeling’ and academic studies ‘carried out a long time ago’ more so than analytics. This approach was common among participants who experienced data availability as a barrier.

Six participants described how concerns around data accuracy sometimes led to considerable resources being used to determine the ‘correct’ data, which stopped more advanced analytical work. In some cases, participants stopped their use of analytics as part of decision-making due to perceived data inaccuracies, and again relied more on expert opinion. More regular and confident analytics users rarely communicated data availability and accuracy as barriers to use of analytics.

Data linkage

Participants describing concerns around data quality and difficulties they had faced accessing data because of siloed data systems or poor working relationships also reflected how projects that link patient records stored across data systems could help overcome these barriers. Without linkage, data were seen as being often disconnected and stored across siloed data systems that ‘don’t talk to each other’. For example, one commissioner described linking NHS and publicly available data on area-level deprivation to inform their decision-making in this example, prompting them to tailor services to different population groups:

We looked at primary care data … then we looked at some acute data, and we managed to link the acute and primary care data. [After linking with publicly available deprivation data] what we ended up with was six very different projects, so not this blanket one size fits all. (ID013, health commissioner)

They described how this linked data enabled them to see the ‘fuller picture’ of service use for residents who accessed care across organisational boundaries. As a result, they felt more able to holistically understand health needs and more efficiently make strategic health and care decisions. However, they felt unable to make decisions that considered residents’ individual social circumstances or social care use, as local authority records (containing such information) did not contain NHS numbers. NHS numbers were seen as necessary enablers of data linkage:

[With] our local authority data, unfortunately, they didn’t use NHS number at all. So normally where you might get say a 65% to 70% match, or even a 50%/60% … we had nothing … the local authority data could have added value. (ID013, health commissioner)

This participant was fairly exceptional as they conducted their own linkage, and other participants did not currently have access to data linked across services. Most participants expressed a positive view of the potential for data linkage to help them understand needs and inform strategic health and care decision-making. Without linked data, participants made decisions with incomplete data that were ‘heavily caveated’ and evaluated, or again sought alternative information. A handful of participants were setting up data systems that linked records across health and care organisations to enable leaders’ access to linked data.