Simplify access to health care

Context

Kaiser’s digital presence was not one site. Kp.org carried the member and plan-shopping experience, thrive.kp.org carried wellness content, and each had grown its own header and footer with no shared logic for signed-in versus signed-out. Underneath both sat eight regions with their own versions of the same content.

There was a lot to choose from. My focus was the content. What exists, what it should be called, who gets to publish it, and what happens when the site has to tell every member something urgent at once.

Act 1: A content model for the filing cabinet

The Forms and Publications section was regionalized and inconsistent, with no shared model underneath. Before anyone could design a better structure, someone had to say what was actually in there. I scraped every document and link from the Northern California page, separated real content from header and footer noise, then loaded the unique items into Airtable so I could query for patterns.

The Newsletters and articles page had sixty-six links, of which fifty-eight were the site’s own navigation, repeated. Eight were content, and three of those were not newsletters. The page was almost entirely scaffolding.

Org chart as information architecture

Moving from information architecture based on an organizational chart to a content model that could inform publishing and CMS going forward. The model lists each document once. Language is not a level in it, and neither is region. The Health Information Exchange opt-in, which the audit had found listed twice under two different names, resolves to a single node. The structure follows what a member came for, not how the files were stored.

Data requests

The inventory covered Northern California only. I requested the available analytics for all eight regions, July 2015 through February 2016, to find out whether one region’s page held everywhere. It held. Across sixty-four region-months, newsletters never broke four percent of category traffic. Forms and plan-services publications carried nearly all of it, and both spiked every January at open enrollment. In Georgia, plan services nearly quintupled.

Forms and Publications: monthly article visits

Newsletters never broke four percent of category traffic. Plan services nearly tripled in January, at open enrollment.

• Forms
• Plan services
• Newsletters

View the data

Month	Forms	Plan services	Newsletters
Jul 2015	5,688	4,598	316
Aug 2015	6,544	4,521	336
Sep 2015	5,529	4,185	336
Oct 2015	5,656	4,207	364
Nov 2015	5,422	3,833	285
Dec 2015	4,997	4,109	197
Jan 2016	6,413	11,570	282
Feb 2016	7,533	11,801	350

Source: Webtrends data request, kp.org, Northern California July 2015 – February 2016

The audit was a pilot on one regional page, so I recommended repeating it across all eight before committing to a final architecture. Separately, the research team ran a card sort on the same area. More than half of participants split forms from informational resources unprompted. Members arrived independently at the structure the inventory implied.

Act 2: A footer with a rule behind it

The footer is where every team in a large organization goes when they have been told no somewhere else. Kaiser’s was a flat list of thirteen links with no sections, and the future-state version had to serve two properties, eight regions, and both authentication states.

I wrote a content strategy for the global components first, so the footer had something to be measured against instead of a queue of requests. Three rules did the work. Prioritize on historical usage rather than on who asks loudest. Personalize before and after login, by region. And for anything proposed, answer one question: is this popular, required, or are we just looking to promote it?

Scroll and heat maps of the Thrive home page settled the sizing argument. Most of the page went unseen on the way down, which turned the question from how many links would fit into how few could be justified. The recommendation was four columns and about thirteen links, sectioned, with the seal and the broker and employer links hidden from signed-in members because a member is not shopping for a plan.

Act 3: Alerts in need of governance

My second assignment was the site-wide alert system, the Important notices component used for everything from flu season to wildfires. I audited what it was carrying and who was putting it there. On the homepage capture I took into the review deck, the component held nine notices at once, published by four different groups, with no order, no priority, and no way to tell which one mattered.

To discovery what the historical patterns were, I pulled the full publishing history out of the legacy CMS to find the causes. Roughly thirty site-wide notices between September 2014 and March 2016. Ebola, the 2015 measles outbreak, wildfires, a North Bay earthquake, a data breach, labor actions, the Supreme Court’s ACA ruling. At least fourteen different publishers, and no shared rule for scope, priority, or when anything came down.

One field hinted at a framework. The CMS had an expiration date, and every single record had it empty. Nothing was ever scheduled to leave. Notices came down when somebody remembered, which meant the page recorded what each team lobbied most recently about rather than what a member needed to know. No amount of visual hierarchy fixes that. An empty field on thirty records is a governance gap, and it is the reason this act is about governance and not about design.

The click data showed what it cost. One live link read “Click here for more information.” It drew 896 clicks and nobody could say what for.

The same panel, eight weeks apart

Flu was the most clicked notice on kp.org in Q4 2015, more than every other notice combined. By January and February taxes had overtaken it. The panel had no way to respond to that, because nothing in it could expire, reorder, or step down.

• Flu vaccine notice
• 2015 tax notice

View the data

Notice	Q4 2015	Jan–Feb 2016
Flu vaccine	8,895	1,219
2015 taxes	~2,400	6,271

Source: kp.org alerts guidelines deck, “Categories & metrics”. Three of these four figures are exact and appear in the record. The Q4 tax number was only ever drawn as a bar, never written down, so it is shown dashed and marked with a tilde rather than given false precision. The deck tracked the Social Security number, virus, and regional news notices over the same two periods; those are left out here for the same reason.

So I proposed three tiers. P1 for emergencies, site-wide and above everything, with little or no advance notice. P2 for closures and service interruptions, site-wide, with normal lead time and a mandatory expiry. P3 for advance-notice and seasonal items, held in the Important notices panel instead of commanding the page. Each tier fixed a placement, a timing rule, an expiry, and a named approver.

Alerts were never only a content problem. Thirteen groups had standing in what got published, from the issues and crisis team and national and regional communications to legal and regulatory, the member service contact center, and the physicians and medical groups. The deck set out their guiding principles. Three of them mattered here: deliver care first and treat financial concerns as secondary; be a compassionate, responsible provider and employer; and share information quickly and openly.

That third principle is the one that bites. Sharing quickly and openly is in direct tension with an approval chain, and thirteen groups had never had to reconcile the two. Naming that tension was the deliverable. Reaching alignment on it would be theirs going forward.

Close

My piece of the migration mapped what actually sat behind Forms and Publications, argued a global footer down to something a member could use and a health system would approve, and turned a decade of ad hoc notices into a system with tiers, expiry dates, and an owner for each.

The tiers went into the backlog as written requirements to inform a criteria for judging what a notification is and where it belongs. It was also the project that turned an unplanned detour, by way of a hospitality client at Razorfish, into a career direction. The one that led to PayPal.